Question and answer method, system and equipment based on intelligent cockpit and medium

Abstract

The invention provides a question answering method, system and device based on an intelligent cockpit and a medium, the method is applied to a question answering system, a rewriting sub-module optimizes question information input by a user to obtain target question information, then a tool calling sub-module carries out intention recognition on the target question information to determine tool calling information of the target question information, and the tool calling information is sent to a server; meanwhile, the personalized recommendation sub-module performs personalized recommendation to determine tool parameter information of the target question information, and then the task sub-module calls a tool from a tool set according to the tool calling information and an arrangement mode corresponding to the target question information, executes the tool according to the tool parameter information, obtains a question and answer result of the question information input by the user, and sends the question and answer result to the user. Automatic rewriting, intention recognition, personalized recommendation, tool arrangement and tool calling are carried out on the question sentences of the user through the multiple modules to generate the question and answer result, the accuracy and efficiency of question and answer are improved, the personalized experience of the user is enhanced, and intelligent, personalized and efficient interaction service is achieved.

Description

Technical Field

[0001] The present application relates to the field of intelligent question-answering technology, and in particular to a question-answering method based on an intelligent cockpit, a question-answering system based on an intelligent cockpit, an electronic device, and a computer-readable storage medium. Background Art

[0002] Semantic understanding is one of the core tasks of Natural Language Processing (NLP), which aims to enable the in-vehicle system to understand the language instructions of the driver and passengers in real time and accurately, thereby providing a more intelligent interactive experience. With the rapid development of artificial intelligence technology, semantic understanding plays a key role in in-vehicle systems, especially in scenarios such as voice assistants, navigation systems, entertainment control, and driving assistance. Semantic understanding technology helps the system better interpret user intentions and improve driving experience and safety.

[0003] At present, the semantic understanding method for smart cockpits is the traditional pipeline semantic understanding method, which gradually parses user questions by calling text classification models, entity recognition models, and dialogue management modules in sequence to achieve semantic understanding. However, this chain processing method not only involves the series connection of multiple models, which increases the complexity and cost of the system, but also has certain limitations in understanding ability, making it difficult to cope with complex and changing user needs. The existing technology also involves directly processing user questions through an integrated agent, but this technical solution relies on a super-large-scale model with hundreds of billions of parameters, which not only brings high computing resource consumption, but also in resource-constrained scenarios such as smart cockpits, the response speed and efficiency of the model become bottlenecks, limiting its widespread promotion in practical applications. Summary of the invention

[0004] The embodiments of the present application provide a question-answering method, system, device and medium based on a smart cockpit to solve the problems of high cost, low efficiency and insufficient understanding ability of the existing semantic understanding method based on the smart cockpit.

[0005] The embodiment of the present application discloses a question-answering method based on a smart cockpit, which is applied to a question-answering system. The system includes a decision module and a task module connected to the decision module. The decision module includes a rewriting submodule, a tool calling submodule, and a personalized recommendation submodule. The task module includes a tool set and multiple task submodules. The method includes:

[0006] Obtaining user input question information through the rewriting submodule, rewriting the question information, obtaining target question information corresponding to the question information, and sending the target question information to the tool calling submodule and the personalized recommendation submodule respectively;

[0007] The target question information is received by the tool calling submodule, and the intention of the target question information is recognized to obtain the tool calling information corresponding to the target question information, and the tool calling information is sent to the task submodule corresponding to the target question information;

[0008] The target question information is received through the personalized recommendation submodule, and personalized analysis is performed on the target question information to obtain tool parameter information corresponding to the target question information, and the tool parameter information is sent to the task submodule corresponding to the target question information;

[0009] The tool calling information and the tool parameter information are received through the task submodule, an arrangement method for the target question information is determined according to the tool calling information, a tool corresponding to the target question information is called from the tool set according to the arrangement method, and the tool is executed according to the tool parameter information to obtain a question and answer result corresponding to the question information.

[0010] Optionally, the obtaining user input question information through the rewriting submodule, and rewriting the question information to obtain target question information corresponding to the question information includes:

[0011] The rewriting submodule obtains user input question information and the user's historical conversation information, performs semantic analysis on the question information, obtains pronoun information in the question information, and rewrites the pronoun information according to the historical conversation information to obtain target question information corresponding to the question information.

[0012] Optionally, the performing intent recognition on the target question information through a tool calling submodule to obtain tool calling information corresponding to the target question information includes:

[0013] The tool calling submodule performs intent recognition on the target question information to obtain an intent recognition result of the target question information, and matches the intent recognition result with preset tool description information to obtain tool calling information corresponding to the target question information.

[0014] Optionally, the performing personalized analysis on the target question information by the personalized recommendation submodule to obtain tool parameter information corresponding to the target question information includes:

[0015] Performing demand analysis on the target question information by the personalized recommendation submodule to obtain a demand analysis result of the target question information;

[0016] If the demand analysis result indicates that the target question information has a personalized recommendation demand, the user's historical conversation information is obtained through the personalized recommendation submodule, and a personalized analysis is performed on the user based on the historical conversation information to obtain a personalized tag of the user, and the personalized tag is used as the tool parameter information corresponding to the target question information.

[0017] Optionally, the target question information corresponds to a plurality of tools, and the determining, by the task submodule, the arrangement mode corresponding to the target question information according to the tool calling information includes:

[0018] The task submodule determines multiple tools corresponding to the target question information according to the tool calling information, and determines an arrangement mode for the multiple tools according to the characteristics of the multiple tools, wherein the arrangement mode is active arrangement or passive arrangement.

[0019] Optionally, calling the tool corresponding to the target question information from the tool set according to the arrangement mode by the task submodule includes:

[0020] If the arrangement mode is active arrangement, the task submodule is used to screen and arrange the plurality of tools to obtain a task completion path for the target question information, and the tool corresponding to the target question information is called from the tool set according to the task completion path;

[0021] If the arrangement mode corresponding to the user question is passive arrangement, the tool corresponding to the target question information is called from the tool set through the task submodule according to a preset task path.

[0022] Optionally, the task submodules correspond to different processing types, and the sending of the tool calling information to the task submodule corresponding to the target question information through the tool calling submodule includes:

[0023] Performing intent recognition on the target question information through the tool calling submodule to obtain an intent recognition result of the target question information, determining a processing type corresponding to the target question information according to the intent recognition result, and sending the tool calling information to a task submodule corresponding to the target question information according to the processing type;

[0024] The sending the tool parameter information to the task submodule corresponding to the target question information through the personalized recommendation submodule includes:

[0025] The personalized recommendation submodule performs intent recognition on the target question information to obtain an intent recognition result of the target question information, determines a processing type corresponding to the target question information according to the intent recognition result, and sends the tool parameter information to a task submodule corresponding to the target question information according to the processing type.

[0026] The embodiment of the present application also provides a question-answering system based on a smart cockpit, the system comprising a decision module and a task module connected to the decision module, the decision module comprising a rewriting submodule, a tool calling submodule and a personalized recommendation submodule, the task module comprising a tool set and a plurality of task submodules; wherein,

[0027] The rewriting submodule is used to obtain the question information input by the user, rewrite the question information, obtain the target question information corresponding to the question information, and send the target question information to the tool calling submodule and the personalized recommendation submodule respectively;

[0028] The tool calling submodule is used to receive the target question information, perform intent recognition on the target question information, obtain tool calling information corresponding to the target question information, and send the tool calling information to the task submodule corresponding to the target question information;

[0029] The personalized recommendation submodule is used to receive the target question information, perform personalized analysis on the target question information, obtain tool parameter information corresponding to the target question information, and send the tool parameter information to the task submodule corresponding to the target question information;

[0030] The task submodule is used to receive the tool calling information and the tool parameter information, determine the arrangement method for the target question information according to the tool calling information, call the tool corresponding to the target question information from the tool set according to the arrangement method, and execute the tool according to the tool parameter information to obtain the question and answer result corresponding to the question information.

[0031] Optionally, the tool calling submodule is specifically used to obtain user input question information and the user's historical conversation information, and perform semantic analysis on the question information to obtain pronoun information in the question information, and rewrite the pronoun information according to the historical conversation information to obtain target question information corresponding to the question information.

[0032] Optionally, the tool calling submodule is specifically used to perform intent recognition on the target question information, obtain an intent recognition result of the target question information, and match the intent recognition result with preset tool description information to obtain tool calling information corresponding to the target question information.

[0033] Optionally, the personalized recommendation submodule is specifically used to: perform demand analysis on the target question information to obtain a demand analysis result of the target question information;

[0034] If the demand analysis result indicates that the target question information has a personalized recommendation demand, historical conversation information of the user is obtained, and personalized analysis is performed on the user based on the historical conversation information to obtain a personalized tag of the user, and the personalized tag is used as tool parameter information corresponding to the target question information.

[0035] Optionally, the target question information corresponds to multiple tools, and the task submodule is specifically used to determine the multiple tools corresponding to the target question information according to the tool calling information, and determine the orchestration method for the multiple tools according to the characteristics of the multiple tools, and the orchestration method is active orchestration or passive orchestration.

[0036] Optionally, the task submodule is further used to:

[0037] If the arrangement mode is active arrangement, the plurality of tools are screened and arranged to obtain a task completion path for the target question information, and the tool corresponding to the target question information is called from the tool set according to the task completion path;

[0038] If the arrangement mode corresponding to the user question is passive arrangement, the tool corresponding to the target question information is called from the tool set according to a preset task path.

[0039] Optionally, the task submodules correspond to different processing types; wherein,

[0040] The tool calling submodule is further used to perform intent recognition on the target question information, obtain an intent recognition result of the target question information, determine a processing type corresponding to the target question information according to the intent recognition result, and send the tool calling information to the task submodule corresponding to the target question information according to the processing type;

[0041] The personalized recommendation submodule is also used to perform intent recognition on the target question information, obtain the intent recognition result of the target question information, determine the processing type corresponding to the target question information according to the intent recognition result, and send the tool parameter information to the task submodule corresponding to the target question information according to the processing type.

[0042] The embodiment of the present application also discloses an electronic device, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus;

[0043] The memory is used to store computer programs;

[0044] The processor is used to implement the method described in the embodiment of the present application when executing the program stored in the memory.

[0045] The embodiment of the present application also discloses a computer-readable storage medium having instructions stored thereon, which, when executed by one or more processors, enables the processors to execute the method described in the embodiment of the present application.

[0046] Compared with the prior art, the embodiments of the present application have the following advantages:

[0047] In an embodiment of the present application, a question-answering system based on a smart cockpit includes a decision module and a task module connected to the decision module. The decision module includes a rewriting submodule, a tool calling submodule and a personalized recommendation submodule. The task module includes a toolset and multiple task submodules. The rewriting submodule optimizes the user input question information to obtain the target question information. Then the tool calling submodule performs intent recognition on the target question information to determine the tool calling information of the target question information. At the same time, the personalized recommendation submodule performs personalized recommendation to determine the tool parameter information of the target question information. Then the task submodule calls the tool from the toolset according to the tool calling information according to the orchestration mode corresponding to the target question information, and executes the tool according to the tool parameter information to obtain the question-answering result of the user input question information. Through the collaborative work of multiple modules, the user's question is automatically rewritten, intent recognized, personalized recommended, tool orchestrated and tool called to generate the question-answering result, thereby improving the accuracy and efficiency of the question-answering, enhancing the user's personalized experience, and realizing intelligent, personalized and efficient interactive services. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0049] Figure 1 It is a flow chart of an existing semantic understanding method provided by an embodiment of the present application;

[0050] Figure 2is a flow chart of another existing semantic understanding method provided by an embodiment of the present application;

[0051] Figure 3 is a flowchart of a question-and-answer method based on a smart cockpit provided in an embodiment of the present application;

[0052] Figure 4 It is a schematic diagram of a scenario in which a task submodule actively arranges and calls a tool provided in an embodiment of the present application;

[0053] Figure 5 is a scenario diagram of a question-and-answer based on a smart cockpit provided in an embodiment of the present application;

[0054] Figure 6 is a structural block diagram of a question-answering system based on a smart cockpit provided in an embodiment of the present application;

[0055] Figure 7 is a block diagram of an electronic device provided in an embodiment of the present application;

[0056] Figure 8 It is a schematic diagram of a computer-readable storage medium provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0058] Semantic understanding is one of the core tasks of Natural Language Processing (NLP), which aims to enable the in-vehicle system to understand the language instructions of the driver and passengers in real time and accurately, thereby providing a more intelligent interactive experience. With the rapid development of artificial intelligence technology, semantic understanding plays a key role in in-vehicle systems, especially in scenarios such as voice assistants, navigation systems, entertainment control, and driving assistance. Semantic understanding technology helps the system better interpret user intentions and improve driving experience and safety.

[0059] As an example, see Figure 1, showing a flow chart of an existing semantic understanding method provided by an embodiment of the present application. For the semantic understanding task of user questions, the natural language understanding module is first used to understand the user input question, identify the user's intention and extract relevant entity information, and then the dialogue management module determines the next response based on the output of the natural language understanding module, and asks to obtain the missing information of the user's question, or makes a decision based on the user's intention and context information. Finally, the natural language generation module calls the external tool and generates a consistent result corresponding to the user's question based on the speech template. In this method, the semantic understanding task is decomposed into a series of steps and executed sequentially. This chain processing method not only involves the series connection of multiple models, which increases the complexity and cost of the system, but also has certain limitations in understanding ability, making it difficult to cope with complex and changeable user needs.

[0060] As another example, refer to Figure 2 , showing a flow chart of another existing semantic understanding method provided by an embodiment of the present application. The prior art also relates to a large model method based on a single intelligent agent, which processes user questions and outputs results through a single intelligent agent module and a tool set. The intelligent agent module specifically guides the LLM (Large Language Model) large model to semantically understand the user questions through prompt words, and then calls tools from the tool set to process the requirements corresponding to the user questions according to the model output results, and finally generates output results corresponding to the user questions. However, this method relies on a very large-scale model with hundreds of billions of parameters, which not only brings high computing resource consumption, but also in resource-constrained scenarios such as smart cockpits, the response speed and efficiency of the model become bottlenecks, limiting its widespread promotion in practical applications.

[0061] In this regard, the advantages of the present application over the prior art are that, in an embodiment of the present application, the rewriting submodule optimizes the user input question information to obtain the target question information, and then the tool calling submodule performs intent recognition on the target question information to determine the tool calling information of the target question information, and at the same time the personalized recommendation submodule performs personalized recommendation to determine the tool parameter information of the target question information, and then the task submodule calls the tool from the tool set according to the orchestration mode corresponding to the target question information based on the tool calling information, and executes the tool according to the tool parameter information to obtain the question and answer results of the user input question information, and through the collaborative work of multiple modules, the user's questions are automatically rewritten, intent recognized, personalized recommended, tool orchestrated and tool called to generate question and answer results, thereby improving the accuracy and efficiency of question and answer, and enhancing the user's personalized experience, realizing intelligent, personalized and efficient interactive services.

[0062] Reference Figure 3, shows a flowchart of the steps of a question-answering method based on a smart cockpit provided in an embodiment of the present application, the method is applied to a question-answering system, the system includes a decision module and a task module connected to the decision module, the decision module includes a rewriting submodule, a tool calling submodule and a personalized recommendation submodule, the task module includes a tool set and multiple task submodules, and specifically may include the following steps:

[0063] Step 301: obtaining user input question information through the rewriting submodule, rewriting the question information, obtaining target question information corresponding to the question information, and sending the target question information to the tool calling submodule and the personalized recommendation submodule respectively;

[0064] In the embodiment of the present application, the question-answering system can be installed in a vehicle, that is, as an on-board system. The question-answering system includes a decision module and a task module. The decision module is responsible for understanding the question information input by the user and determining the user's needs. The task module is used to arrange and call the tools corresponding to the user's questions according to the output information of the decision module, and output the question-answering results corresponding to the user's questions. Specifically, the decision module includes a rewriting submodule, a tool calling submodule, and a personalized recommendation submodule. The rewriting submodule is used to optimize the question information to generate more complete target question information. The tool calling submodule is used to determine the tool that needs to be called for the target question information. The personalized recommendation submodule is used to output the user's personalized label as a tool parameter; the task module includes a tool set and multiple task submodules. The tool set contains multiple tools for completing the specific needs of the user. The task submodule is used to further arrange and call the tools to complete the user's needs and output the question-answering results corresponding to the user's input question information.

[0065] In the specific implementation process, the user inputs question information, which can be in various forms such as text information, voice information, and picture information. For example, the user inputs "Please help me check the weather in Beijing" through voice input or text input of the vehicle-mounted device, or the user inputs a picture and writes "Help me book a room at this hotel", and then obtains the question information input by the user through the rewriting submodule, rewrites the question information, optimizes the question information to obtain the target question information that accurately represents the user's needs, and then sends the optimized target question information to the tool calling submodule and the personalized recommendation submodule respectively, to provide high-quality input for subsequent modules. In the embodiment of the present application, the question information input by the user is optimized into a clearer and more complete target question through the rewriting submodule, avoiding the understanding deviation caused by the irregular user input, and avoiding the problems of unclear expression, irregular grammar, incomplete information, etc. in the question input by the user.

[0066] In a preferred embodiment of the present application, the step of obtaining user input question information through the rewriting submodule, and rewriting the question information to obtain target question information corresponding to the question information includes:

[0067] The rewriting submodule obtains user input question information and the user's historical conversation information, performs semantic analysis on the question information, obtains pronoun information in the question information, and rewrites the pronoun information according to the historical conversation information to obtain target question information corresponding to the question information.

[0068] In the embodiment of the present application, the rewriting submodule also uses the user's historical conversation information as auxiliary information to understand the question information currently input by the user, and then rewrites the question information according to the historical conversation information to obtain the target question information. Specifically, the rewriting submodule identifies the pronoun information in the question information, such as this, he, that, etc., and then rewrites the pronouns in the question information according to the historical conversation information. For example, the user mentioned in the previous historical conversation "I want to listen to singer A's song", and then rewrites the current question information "play his song" to "play singer A's song", and then outputs the optimized target question information. The embodiment of the present application replaces pronouns with specific entities with the assistance of semantic analysis and historical conversation information, thereby making the question clearer.

[0069] In a preferred embodiment of the present application, multiple task sub-modules in the task module are pre-set to correspond to different processing types, and then the question information and the corresponding tool call information and tool parameter information are sent to the corresponding task sub-module according to the processing type of the question information input by the user, ensuring that different task sub-modules are flexibly called according to user needs to support diverse user needs.

[0070] Specifically, by performing intent recognition on the target question information, the processing type corresponding to the target question information is determined, and then the target question information and the corresponding tool call information and tool parameter information are sent to the corresponding task submodule according to the processing type. For example, a travel planning task submodule and an emotional soothing task submodule are pre-set. If the user inputs the question "Please help me plan a travel route to Beijing" and identifies the user's intention as travel planning, it will be sent to the travel planning task submodule for processing. If the user inputs the question "I'm so tired" and identifies the user's intention as emotional soothing, it will be sent to the emotional soothing task submodule for processing. The embodiment of the present application ensures that different task submodules are flexibly called according to user needs and supports diverse user needs.

[0071] Step 302: receiving the target question information through the tool calling submodule, performing intent recognition on the target question information, obtaining tool calling information corresponding to the target question information, and sending the tool calling information to the task submodule corresponding to the target question information;

[0072] In an embodiment of the present application, the tool calling submodule performs intent recognition on the target question information, clarifies the user's needs, and then determines the tools required to meet the user's needs, and generates tool calling information corresponding to the target question information. Specifically, the tool calling submodule analyzes the intention of the user's question, such as by performing intent recognition on the target question information to clarify whether the user wants to listen to music, check the weather or make travel plans, and then determines the tools that need to be called according to the intention recognition results, such as music playback tools, weather query tools, ticket booking tools, navigation tools, etc., and generates tool calling information. The tool calling information is used to describe the tools that need to be called, and finally the tool calling information is sent to the corresponding task submodule for the task submodule to perform specific tool calling operations. The embodiment of the present application uses the tool calling submodule to flexibly select and call different tools according to the intention recognition results of the question information to support diverse user needs.

[0073] In a preferred embodiment of the present application, the step of performing intent recognition on the target question information through a tool calling submodule to obtain tool calling information corresponding to the target question information includes:

[0074] The tool calling submodule performs intent recognition on the target question information to obtain an intent recognition result of the target question information, and matches the intent recognition result with preset tool description information to obtain tool calling information corresponding to the target question information.

[0075] In an embodiment of the present application, the tool calling submodule also includes pre-set tool description information, which is a description of the corresponding function of the tool. After the tool calling submodule identifies the user's intention and determines the user's needs, the user needs are matched with the tool description information to determine the tool that meets the user's needs, thereby ensuring that the tool that matches the user's needs can be accurately called. For example, the user inputs the question information "Adjust the temperature to 25 degrees." It is recognized that the user's intention is to adjust the temperature, and the tool corresponding to the tool description information "Adjust the air conditioning temperature" is matched to the "air conditioning control system."

[0076] Step 303: receiving the target question information through the personalized recommendation submodule, performing personalized analysis on the target question information, obtaining tool parameter information corresponding to the target question information, and sending the tool parameter information to the task submodule corresponding to the target question information;

[0077] In the embodiment of the present application, after receiving the target question information, the personalized recommendation submodule will perform personalized analysis on the target question information, determine the user personalized features corresponding to the target question information, such as user preferences, and then generate tool parameter information that matches the user needs and user personalization. The tool parameter information is used to determine the specific use of the tool. The tool parameter information can be used to further refine the use of the tool. For example, if the user inputs the question information as "play music", the personalized recommendation submodule can perform personalized analysis to determine that the user's music preference is rock music, and then generate tool parameter information of the music type of rock music, and then call the music playing tool to play rock music. The embodiment of the present application realizes the linkage and combination of tools and user personalization through the personalized recommendation submodule, thereby providing users with personalized interactive services.

[0078] In a preferred embodiment of the present application, the step of performing personalized analysis on the target question information by the personalized recommendation submodule to obtain tool parameter information corresponding to the target question information includes:

[0079] Performing demand analysis on the target question information by the personalized recommendation submodule to obtain a demand analysis result of the target question information;

[0080] If the demand analysis result indicates that the target question information has a personalized recommendation demand, the user's historical conversation information is obtained through the personalized recommendation submodule, and a personalized analysis is performed on the user based on the historical conversation information to obtain a personalized tag of the user, and the personalized tag is used as the tool parameter information corresponding to the target question information.

[0081] In the embodiment of the present application, after receiving the target question information, the personalized recommendation submodule will first determine whether the question information input by the user has a need for personalized recommendation. If there is a need for personalized recommendation, the user will be personalized analyzed based on the user's historical conversation information to obtain the user's personalized features, that is, personalized tags, and then the personalized tags will be used as tool parameter information. In the embodiment of the present application, the tool parameter information is used to make the execution results of the tool more in line with the user's expectations, thereby improving the accuracy of the question and answer results.

[0082] Specifically, the personalized recommendation submodule performs a demand analysis on the target question information, and determines whether there is a personalized recommendation demand based on the demand analysis results. If not, no tool parameter information is generated. If so, the user's historical conversation information is obtained as the basis for personalized analysis. The user is personalized analyzed based on the historical conversation information, and a personalized tag for the user is generated. The personalized tag is used as the tool parameter information corresponding to the target question information. The tool parameter information can then be used to further refine the way the tool is used. For example, if the user inputs the question "What's the weather like today?", no personalized recommendation is required. If the user inputs the question "Play music," there is a demand for personalized recommendation. Through the user's historical conversation records, it is found that the user likes to listen to rock music, and the tag of rock music is used as the tool parameter information.

[0083] Step 304: receiving the tool calling information and the tool parameter information through the task submodule, determining an arrangement method for the target question information according to the tool calling information, calling a tool corresponding to the target question information from the tool set according to the arrangement method, and executing the tool according to the tool parameter information to obtain a question and answer result corresponding to the question information.

[0084] In an embodiment of the present application, after receiving the tool calling information and tool parameter information, the task submodule can first clarify the tool that needs to be called for the target question information based on the tool calling information, and further determine the arrangement method for the target question information based on the tool calling information. The arrangement method is used to determine the selection of the tool corresponding to the target question information and the calling order of the tools, and then call the tool from the tool set according to the arrangement method, execute the tool according to the tool parameter information, and output the question and answer result corresponding to the question information.

[0085] In the specific implementation process, the target question information may correspond to multiple tools. If multiple tools need to be called to meet the user's needs, the task submodule will determine the tool arrangement method based on the characteristics of the tool after obtaining the multiple tools corresponding to the target question information through the tool calling information. The arrangement method includes active arrangement and passive arrangement. Active arrangement means that the task submodule will further select the multiple tools corresponding to the target question information and determine the calling order of the tools. Passive arrangement means that the task submodule directly calls the tool according to the tool calling information without performing additional arrangement.

[0086] As an example, if the multiple tools corresponding to the target question information have a fixed calling order and all need to be called, the task submodule adopts a passive arrangement method. If the multiple tools corresponding to the target question information do not have a fixed calling order and are not required to be called, the task submodule can take an active arrangement method. For example: if the task submodule receives the target question information "I am so tired, please lower the air conditioner temperature to 22 degrees and play music", the corresponding tool call information is the air conditioner adjustment tool and the music playing tool, then the task submodule will call the air conditioner adjustment tool to lower the temperature, and call the music playing tool to play music, and perform passive fixed arrangement; if the task submodule receives the target question information "I am so tired", the corresponding tool call information is the language soothing tool, the music playing tool, the navigation tool, the lighting adjustment tool and the air conditioner adjustment tool, the task submodule calls the language soothing tool to output the soothing voice, and further calls the navigation tool to output the recommended food navigation information, and performs active and flexible arrangement.

[0087] In a preferred embodiment of the present application, calling the tool corresponding to the target question information from the tool set according to the arrangement method by the task submodule includes:

[0088] If the arrangement mode is active arrangement, the task submodule is used to screen and arrange the plurality of tools to obtain a task completion path for the target question information, and the tool corresponding to the target question information is called from the tool set according to the task completion path;

[0089] If the arrangement mode corresponding to the user question is passive arrangement, the tool corresponding to the target question information is called from the tool set through the task submodule according to a preset task path.

[0090] In an embodiment of the present application, if the task submodule adopts a passive orchestration method, the tool corresponding to the target question information is called from the tool set according to the pre-set task path for the tool. For example, if the user inputs the question "Recommend a takeaway and help me place an order." The corresponding called tools are a food query tool and a payment tool. Then, according to the preset task completion path, the food recommendation tool is first called to query recommended takeaway merchants, and then the payment tool is further called to place an order for the takeaway.

[0091] If the task submodule adopts the active arrangement method, multiple tools are screened and arranged to generate a task completion path for the target question information. According to the task completion path, the tool corresponding to the target question information is called from the tool set; Figure 4, which is a schematic diagram of a scenario in which a task submodule provided in an embodiment of the present application actively arranges and calls tools. For example, if the user inputs a question sentence "I am so tired", it is sent to the emotional care task submodule for processing. The tool call information is a language soothing tool, a music playing tool, a food recommendation tool, a lighting adjustment tool, and an air conditioning adjustment tool. The task submodule screens and arranges the tools, and generates a task completion path of a language soothing tool-music playing tool-food recommendation tool. Then, according to the task completion path, the language soothing tool is called to output a soothing voice, and then the music playing tool is called to play music, and the food recommendation tool is further called to find food, and finally the corresponding question and answer results are summarized and output.

[0092] Optionally, the task submodule can be pre-set to adopt a fixed active orchestration mode or a passive orchestration mode according to the processing type of the task submodule, so that when the task submodule receives the tool calling information, it directly calls the tool from the tool set according to the preset task path and executes it to output the result; or the task submodule will perform active orchestration after receiving the tool calling information, generate a dynamic task completion path, and then call the tool through the task completion path, so that for some common and fixed task scenarios (such as navigation, music playback, etc.), the fixed orchestration method can quickly respond to user needs and improve user experience, and actively orchestrate some complex task scenarios (such as emotional care, etc.) to meet diverse user needs.

[0093] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present application, the following examples are used for exemplary description:

[0094] Reference Figure 5 , showing a scenario diagram of a question-and-answer system based on a smart cockpit provided in an embodiment of the present application. For the question-and-answer system based on a smart cockpit, the rewriting submodule, tool calling submodule, personalized recommendation submodule included in the decision module and the multiple task submodules included in the task module are all intelligent agent modules based on the LLM big model, and the LLM big model corresponding to each module can adopt a big model with different parameters according to the complexity of task processing, and the question-and-answer system can also include an example library.

[0095] In the specific implementation process, when the rewriting submodule obtains the user question, it will first obtain the prompt word corresponding to the user question and the user's historical conversation from the sample library, and input it into the large model to drive the LLM large model to rewrite the user question, obtain the optimized target user question information, and then input the target user question information into the tool calling submodule and the personalized recommendation submodule respectively; then the tool calling submodule obtains the prompt word corresponding to the target user question information from the sample library and inputs it into the target user question information in combination with the toolset description, so as to drive the LLM large model to determine the calling tool corresponding to the target user question information and send it to the corresponding task. At the same time, the personalized recommendation submodule obtains the prompt words corresponding to the target user's question information from the example library and drives the LLM big model in combination with the personalized question examples to determine whether the target user's question information has personalized recommendation needs. If so, the user's personalized label is further determined by the LLM big model as the tool parameter information and sent to the corresponding task submodule; finally, after receiving the target question information and the corresponding tool call information and tool parameter information, the task submodule will arrange the tools corresponding to the target question information, and then call the corresponding tools from the toolset and execute the tools according to the tool parameter information, and output the question and answer results corresponding to the user's question.

[0096] In addition, the LLM large model corresponding to different modules can pre-set the corresponding output length threshold, and then limit the output result length of the LLM large model according to the preset output length threshold.

[0097] It should be noted that, for the method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited by the described order of actions, because according to the embodiments of the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.

[0098] Reference Figure 6 , shows a structural block diagram of a question-answering system based on an intelligent cockpit provided in an embodiment of the present application, the system includes a decision module and a task module connected to the decision module, the decision module includes a rewriting submodule, a tool calling submodule and a personalized recommendation submodule, the task module includes a tool set and multiple task submodules; wherein,

[0099] The rewriting submodule is used to obtain the question information input by the user, rewrite the question information, obtain the target question information corresponding to the question information, and send the target question information to the tool calling submodule and the personalized recommendation submodule respectively;

[0100] The tool calling submodule is used to receive the target question information, perform intent recognition on the target question information, obtain tool calling information corresponding to the target question information, and send the tool calling information to the task submodule corresponding to the target question information;

[0101] The personalized recommendation submodule is used to receive the target question information, perform personalized analysis on the target question information, obtain tool parameter information corresponding to the target question information, and send the tool parameter information to the task submodule corresponding to the target question information;

[0102] The task submodule is used to receive the tool calling information and the tool parameter information, determine the arrangement method for the target question information according to the tool calling information, call the tool corresponding to the target question information from the tool set according to the arrangement method, and execute the tool according to the tool parameter information to obtain the question and answer result corresponding to the question information.

[0103] In one embodiment of the present application, the tool calling submodule is specifically used to obtain user input question information and the user's historical conversation information, and perform semantic analysis on the question information to obtain pronoun information in the question information, and rewrite the pronoun information according to the historical conversation information to obtain target question information corresponding to the question information.

[0104] In one embodiment of the present application, the tool calling submodule is specifically used to perform intent recognition on the target question information, obtain the intent recognition result of the target question information, and match the intent recognition result with preset tool description information to obtain tool calling information corresponding to the target question information.

[0105] In an embodiment of the present application, the personalized recommendation submodule is specifically used to: perform demand analysis on the target question information to obtain a demand analysis result of the target question information;

[0106] If the demand analysis result indicates that the target question information has a personalized recommendation demand, historical conversation information of the user is obtained, and personalized analysis is performed on the user based on the historical conversation information to obtain a personalized tag of the user, and the personalized tag is used as tool parameter information corresponding to the target question information.

[0107] In one embodiment of the present application, the target question information corresponds to multiple tools, and the task submodule is specifically used to determine the multiple tools corresponding to the target question information according to the tool calling information, and determine the orchestration method for the multiple tools according to the characteristics of the multiple tools, and the orchestration method is active orchestration or passive orchestration.

[0108] In one embodiment of the present application, the task submodule is further used for:

[0109] If the arrangement mode is active arrangement, the plurality of tools are screened and arranged to obtain a task completion path for the target question information, and the tool corresponding to the target question information is called from the tool set according to the task completion path;

[0110] If the arrangement mode corresponding to the user question is passive arrangement, the tool corresponding to the target question information is called from the tool set according to a preset task path.

[0111] In one embodiment of the present application, the task submodules correspond to different processing types; wherein,

[0112] The tool calling submodule is further used to perform intent recognition on the target question information, obtain an intent recognition result of the target question information, determine a processing type corresponding to the target question information according to the intent recognition result, and send the tool calling information to the task submodule corresponding to the target question information according to the processing type;

[0113] The personalized recommendation submodule is also used to perform intent recognition on the target question information, obtain the intent recognition result of the target question information, determine the processing type corresponding to the target question information according to the intent recognition result, and send the tool parameter information to the task submodule corresponding to the target question information according to the processing type.

[0114] As for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0115] In addition, the present application also provides an electronic device, such as Figure 7 As shown, it includes a processor 701, a communication interface 702, a memory 703 and a communication bus 704, wherein the processor 701, the communication interface 702, and the memory 703 communicate with each other through the communication bus 704.

[0116] Memory 703, used for storing computer programs;

[0117] The processor 701 is used to implement the method described in the above embodiment when executing the program stored in the memory 703.

[0118] The communication bus mentioned in the above terminal can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0119] The communication interface is used for communication between the above terminal and other devices.

[0120] The memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0121] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0122] like Figure 8 As shown, in another embodiment provided in the present application, a computer-readable storage medium 801 is also provided, in which instructions are stored. When the instructions are executed by one or more processors, the processors execute the methods described in the above embodiments.

[0123] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0124] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0125] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0126] The above description is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application are included in the protection scope of the present application.

Claims

1. A question-answering method based on an intelligent cockpit, characterized in that: Applied to a question-answering system, the system includes a decision module and a task module connected to the decision module, the decision module includes a rewriting submodule, a tool calling submodule and a personalized recommendation submodule, the task module includes a tool set and multiple task submodules, and the method includes: Obtaining user input question information through the rewriting submodule, rewriting the question information, obtaining target question information corresponding to the question information, and sending the target question information to the tool calling submodule and the personalized recommendation submodule respectively; The target question information is received by the tool calling submodule, and the intention of the target question information is recognized to obtain the tool calling information corresponding to the target question information, and the tool calling information is sent to the task submodule corresponding to the target question information; The target question information is received through the personalized recommendation submodule, and personalized analysis is performed on the target question information to obtain tool parameter information corresponding to the target question information, and the tool parameter information is sent to the task submodule corresponding to the target question information; The tool calling information and the tool parameter information are received through the task submodule, an arrangement method for the target question information is determined according to the tool calling information, a tool corresponding to the target question information is called from the tool set according to the arrangement method, and the tool is executed according to the tool parameter information to obtain a question and answer result corresponding to the question information.

2. The method according to claim 1, characterized in that: The step of obtaining the user input question information through the rewriting submodule, and rewriting the question information to obtain the target question information corresponding to the question information includes: The rewriting submodule obtains user input question information and the user's historical conversation information, performs semantic analysis on the question information, obtains pronoun information in the question information, and rewrites the pronoun information according to the historical conversation information to obtain target question information corresponding to the question information.

3. The method according to claim 1, characterized in that: The performing intent recognition on the target question information through the tool calling submodule to obtain tool calling information corresponding to the target question information includes: The tool calling submodule performs intent recognition on the target question information to obtain an intent recognition result of the target question information, and matches the intent recognition result with preset tool description information to obtain tool calling information corresponding to the target question information.

4. The method according to claim 1, characterized in that: The performing personalized analysis on the target question information by the personalized recommendation submodule to obtain tool parameter information corresponding to the target question information includes: Performing demand analysis on the target question information by the personalized recommendation submodule to obtain a demand analysis result of the target question information; If the demand analysis result indicates that the target question information has a personalized recommendation demand, the user's historical conversation information is obtained through the personalized recommendation submodule, and a personalized analysis is performed on the user based on the historical conversation information to obtain a personalized tag of the user, and the personalized tag is used as the tool parameter information corresponding to the target question information.

5. The method according to claim 1, characterized in that: The target question information corresponds to a plurality of tools, and the determining, by the task submodule, the arrangement mode corresponding to the target question information according to the tool calling information includes: The task submodule determines multiple tools corresponding to the target question information according to the tool calling information, and determines an arrangement mode for the multiple tools according to the characteristics of the multiple tools, wherein the arrangement mode is active arrangement or passive arrangement.

6. The method according to claim 5, characterized in that: The calling of the tool corresponding to the target question information from the tool set according to the arrangement method by the task submodule includes: If the arrangement mode is active arrangement, the task submodule is used to screen and arrange the plurality of tools to obtain a task completion path for the target question information, and the tool corresponding to the target question information is called from the tool set according to the task completion path; If the arrangement mode corresponding to the user question is passive arrangement, the tool corresponding to the target question information is called from the tool set through the task submodule according to a preset task path.

7. The method according to claim 1, characterized in that: The task submodules correspond to different processing types, and the tool calling submodule sends the tool calling information to the task submodule corresponding to the target question information, including: Performing intent recognition on the target question information through the tool calling submodule to obtain an intent recognition result of the target question information, determining a processing type corresponding to the target question information according to the intent recognition result, and sending the tool calling information to a task submodule corresponding to the target question information according to the processing type; The sending the tool parameter information to the task submodule corresponding to the target question information through the personalized recommendation submodule includes: The personalized recommendation submodule performs intent recognition on the target question information to obtain an intent recognition result of the target question information, determines a processing type corresponding to the target question information according to the intent recognition result, and sends the tool parameter information to a task submodule corresponding to the target question information according to the processing type.

8. A question-answering system based on an intelligent cockpit, characterized in that: The system includes a decision module and a task module connected to the decision module, the decision module includes a rewriting submodule, a tool calling submodule and a personalized recommendation submodule, and the task module includes a tool set and multiple task submodules; wherein, The rewriting submodule is used to obtain the question information input by the user, rewrite the question information, obtain the target question information corresponding to the question information, and send the target question information to the tool calling submodule and the personalized recommendation submodule respectively; The tool calling submodule is used to receive the target question information, perform intent recognition on the target question information, obtain tool calling information corresponding to the target question information, and send the tool calling information to the task submodule corresponding to the target question information; The personalized recommendation submodule is used to receive the target question information, perform personalized analysis on the target question information, obtain tool parameter information corresponding to the target question information, and send the tool parameter information to the task submodule corresponding to the target question information; The task submodule is used to receive the tool calling information and the tool parameter information, determine the arrangement method for the target question information according to the tool calling information, call the tool corresponding to the target question information from the tool set according to the arrangement method, and execute the tool according to the tool parameter information to obtain the question and answer result corresponding to the question information.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; The memory is used to store computer programs; The processor is used to implement the method according to any one of claims 1 to 7 when executing the program stored in the memory.

10. A computer-readable storage medium having instructions stored thereon, which, when executed by one or more processors, cause the processors to perform the method according to any one of claims 1 to 7.

Images