Dialogue method, apparatus, system, device, medium, program product and vehicle

Abstract

The application discloses a dialogue method, device, system, equipment, medium, program product and vehicle, and belongs to the technical field of natural language processing. The method comprises the following steps: receiving a dialogue request; determining a dialogue backtracking strategy according to the dialogue request and stored dialogue key information; the dialogue key information comprises dialogue content, dialogue content integrity and dialogue progress information; and outputting target reply content corresponding to the dialogue request according to the dialogue backtracking strategy. When the dialogue request is received, the dialogue backtracking strategy is determined according to the stored dialogue key information, the dialogue backtracking can be performed according to the stored dialogue key information, the possibility of obtaining the reply content again is provided for the user, and the dialogue experience of the user is improved.

Description

Technical Field

[0001] This application relates to the field of natural language processing technology, and more specifically to a dialogue method, apparatus, system, device, medium, program product, and vehicle. Background Technology

[0002] Currently, the in-vehicle infotainment system implements a large-scale question-and-answer function, based on cards for single-round interactions with users, answering their questions and providing more assistance and a better experience. Ideally, the interaction process involves the user asking a question, the voice assistant reading the answer while displaying the text on a card, and the single-round large-scale question-and-answer card exiting with a countdown at the end of the voice broadcast, signifying the end of the dialogue round. After the dialogue ends, the information from that round is discarded; it is not stored and cannot be rewound. If the card exits mid-voice broadcast for any reason, such as the user actively exiting due to other commitments, being exited by other services, or even accidentally exiting, the user will not be able to read or hear all the content on the current card. Because the responses from large-scale models are currently quite long, this mid-interval exit is very common.

[0003] After a large model's question-and-answer card disappears, if a user wants to continue learning about the previous topic, they can only ask the same question again. However, due to the limitations of the large model's generation mechanism, users cannot get a completely identical answer and must listen to the answer again from the beginning. Currently, there is no good solution for this scenario to allow users to revisit the conversation; a lost card cannot be displayed again, and the half of the answer missed last time is forever unavailable. Summary of the Invention

[0004] The purpose of this application is to provide a dialogue method, apparatus, system, device, medium, program product, and vehicle to address the shortcomings of existing technologies where responses cannot be retrieved after a large-scale dialogue session is terminated midway.

[0005] To achieve the above objectives, the first aspect of this application provides a dialogue method, the dialogue method comprising:

[0006] Receive a dialogue request;

[0007] Based on the dialogue request and the stored key dialogue information, a dialogue backtracking strategy is determined; the key dialogue information includes: dialogue content, dialogue content completeness, and dialogue progress information.

[0008] Based on the dialogue backtracking strategy, output the target response content corresponding to the dialogue request.

[0009] Based on the aforementioned technical means, when a dialogue request is received, a dialogue backtracking strategy is determined according to the stored key dialogue information. This allows for dialogue backtracking based on the stored key dialogue information, providing users with the possibility of obtaining the response content again and improving the user's dialogue experience.

[0010] In some feasible embodiments, the dialogue content includes a first response; determining the dialogue backtracking strategy based on the dialogue request and stored key dialogue information includes:

[0011] Determine if there is key information in the dialogue to determine if there is dialogue content that can be used for backtracking;

[0012] Based on the completeness of the dialogue content, determine whether the first response content has been fully generated;

[0013] Based on the dialogue progress information, determine whether the first response content has been completely output;

[0014] The corresponding dialogue backtracking strategy is determined based on whether there is dialogue content for backtracking, whether the first response content has been generated, and whether the first response content has been output.

[0015] Based on the aforementioned technical means, the presence or absence of key dialogue information indicates whether there is dialogue content for backtracking; the completeness of the dialogue content indicates whether the first response content has been generated; and the dialogue progress information indicates whether the first response content has been output. Based on this, the state of the previous round of dialogue can be analyzed, providing a basis for determining the dialogue backtracking strategy.

[0016] In some feasible embodiments, a corresponding dialogue backtracking strategy is determined based on whether there is dialogue content for backtracking, whether the first response content has been generated completely, and whether the first response content has been output completely, including:

[0017] If no dialogue content is available for backtracking, then determine and execute the first dialogue backtracking strategy;

[0018] If there is dialogue content for backtracking, and the first response content has been generated and output, then the second dialogue backtracking strategy is determined and executed.

[0019] If there is dialogue content for backtracking, and the first response content has been generated but not yet output, then the third dialogue backtracking strategy is determined and executed.

[0020] If there is dialogue content for backtracking, and the first response content has not been fully generated, then the fourth dialogue backtracking strategy is determined and executed.

[0021] Based on the aforementioned technical means, the state of the previous round of dialogue is obtained by analyzing key information in the dialogue. Based on the state of the previous round of dialogue, the user's actual need to trigger dialogue backtracking is clarified. Different dialogue backtracking strategies are executed according to different needs to maximize the satisfaction of the user's dialogue backtracking needs and improve the user's dialogue experience.

[0022] In some feasible embodiments, the key dialogue information is stored in the following manner:

[0023] When generating response content at the start of the current round of dialogue, clear the key dialogue information of the previous round of dialogue from the cache, and store the key dialogue information of the current round of dialogue in the cache.

[0024] Based on the aforementioned technical means, key dialogue information is cleared when the response content is generated. If dialogue rewinding is performed, the dialogue questions and responses from the previous round of dialogue have already been obtained from the key dialogue information. Clearing the key dialogue information will not affect the dialogue rewinding, and at the same time, it can ensure that the key dialogue information of the current round of dialogue can be accurately stored.

[0025] In some feasible embodiments, the dialogue backtracking strategy includes: a first dialogue backtracking strategy, a second dialogue backtracking strategy, a third dialogue backtracking strategy, and a fourth dialogue backtracking strategy;

[0026] The first dialogue backtracking strategy includes: starting a new dialogue;

[0027] The second dialogue backtracking strategy includes: obtaining the previous dialogue question and the first response from the key dialogue information;

[0028] Generate supplementary response content based on the dialogue question and the first response content;

[0029] The third dialogue backtracking strategy includes: obtaining the first response content and dialogue progress information from the key dialogue information;

[0030] The fourth dialogue backtracking strategy includes:

[0031] The dialogue questions from the previous round are obtained from the key information of the dialogue, and new response content is generated based on the dialogue questions.

[0032] Based on the aforementioned technical means, starting a new conversation when there is no dialogue content for backtracking allows for a responsive response to the user's dialogue needs, ensuring that every request is answered. When the first response is generated and output, it is determined that the user has consumed all the content from the previous round of dialogue. At this point, the user triggers dialogue backtracking to obtain more content based on previous questions. Therefore, subsequent conversations supplement the responses, providing the user with more needed content and improving the user experience. When the first response is complete, but not fully output, it is determined that the user needs to continue receiving the same content as before. In this case, the first response and dialogue progress information are retrieved, and the content to be output after the strategy is executed is prepared, allowing the user to continue reading or listening to the answers they only partially heard in the previous conversation. When the first response for backtracking is incomplete, new response content is generated based on the dialogue questions, providing complete answers to previously unanswered questions and improving the user experience.

[0033] In some feasible embodiments, the dialogue backtracking strategy includes: a first dialogue backtracking strategy, a second dialogue backtracking strategy, a third dialogue backtracking strategy, and a fourth dialogue backtracking strategy;

[0034] After determining the dialogue backtracking strategy, when the dialogue backtracking strategy is the first dialogue backtracking strategy, the key dialogue information of the new dialogue is stored in the cache.

[0035] When the dialogue backtracking strategy is the second dialogue backtracking strategy, when generating supplementary response content, the key dialogue information of the previous round of dialogue in the cache is cleared, and the dialogue request and the generated supplementary response content are stored in the cache.

[0036] When the dialogue backtracking strategy is the third dialogue backtracking strategy, after obtaining the first reply content from the dialogue key information, the dialogue key information of the previous round of dialogue in the cache is cleared, the dialogue request is stored in the cache, and the first reply content is stored in the cache as the target reply content.

[0037] When the dialogue backtracking strategy is the fourth dialogue backtracking strategy, when generating new response content, the key dialogue information of the previous round of dialogue in the cache is cleared, and the dialogue request and the generated new response content are stored in the cache.

[0038] Based on the above technical means, different dialogue backtracking strategies correspond to different storage methods. When a dialogue backtracking strategy needs to be generated, the key dialogue information of the previous round of dialogue is cleared from the cache when the generation begins, and the dialogue request and the generated content are stored in the cache to ensure that the key dialogue information is updated in real time. When the dialogue backtracking strategy needs to backtrack and output the response content of the previous round, the first response content is stored in the cache as the target response content to ensure that the key dialogue information can be stored normally, and also to ensure the accuracy of the key dialogue information.

[0039] In some feasible embodiments, the step of outputting the target response content corresponding to the dialogue request according to the dialogue backtracking strategy includes:

[0040] The dialogue backtracking strategy is executed to obtain the target response content;

[0041] Based on the target response content, output the target response content according to the output strategy;

[0042] The output strategy includes a voice broadcast strategy and a response content display strategy.

[0043] Based on the aforementioned technical means, the corresponding target response content is obtained after executing the dialogue backtracking strategy. Then, based on the output strategy, different target response content is output according to different user needs, thereby improving the user's dialogue experience.

[0044] In some feasible embodiments, the voice broadcasting strategy includes guidance information output and target response content output, wherein the guidance information output and the target response content output are linked by listening to the end signal of the guidance information output;

[0045] The guidance information includes a new dialogue start prompt voice when the first dialogue backtracking strategy is executed, and an original dialogue continuation prompt voice when the third or fourth dialogue backtracking strategy is executed.

[0046] The target response content output includes: the output of the response content for the new dialogue after executing the first dialogue backtracking strategy;

[0047] Alternatively, it can be the output of supplementary content to the response after executing the second dialogue backtracking strategy;

[0048] Alternatively, after executing the third dialogue backtracking strategy, the output of the first response content can continue based on the dialogue progress information;

[0049] Alternatively, it can be the output of the new response content after executing the fourth dialogue backtracking strategy.

[0050] Based on the aforementioned technical means, guiding information can serve as a bridge in the dialogue process, prompting users on the direction of the conversation. The execution result is output only after the signal indicating the end of the guiding information output has been detected. This avoids interrupting the guiding information output and enhances the completeness of the content the user is listening to. Different output methods are used for the target response content depending on whether it has already been output. Specifically, for responses that were partially output in the previous round, a method is adopted to continue outputting them based on the dialogue progress information, allowing users to continue receiving the half-finished response content from the previous round of dialogue, thus improving the efficiency and experience of obtaining responses during the conversation.

[0051] In some feasible embodiments, after receiving the dialogue request but before determining the dialogue backtracking strategy, the method further includes:

[0052] Identify the intent behind the dialogue request;

[0053] Determine whether the stated intent is the intent of the target vertical domain;

[0054] When the intent is the intent of the target vertical domain, determine whether the dialogue request triggers dialogue backtracking;

[0055] When a dialogue backtracking is triggered, execute the steps to determine the dialogue backtracking strategy.

[0056] Based on the above technical means, after receiving a dialogue request, it can be determined whether to trigger dialogue backtracking based on the identified intent. When dialogue backtracking is triggered, the step of determining the dialogue backtracking strategy is executed, which can avoid executing the step of determining the dialogue backtracking strategy when the dialogue request does not trigger dialogue backtracking, thus avoiding the occupation of system memory.

[0057] In some feasible embodiments, the dialogue method further includes:

[0058] Upon receiving a signal that the conversation has been interrupted, the interruption signal is synchronized to the voice avatar, which then provides a follow-up prompt.

[0059] Based on the aforementioned technical means, when a conversation is interrupted, a voice prompt can help the user recall the conversation that occurred before the interruption, allowing the user to promptly rewind the conversation and obtain the content they want.

[0060] A second aspect of this application provides a dialogue device, the dialogue device comprising:

[0061] The data receiving unit is used to receive dialogue requests;

[0062] The strategy determination unit is used to determine a dialogue backtracking strategy based on the dialogue request and stored key dialogue information; the key dialogue information includes: dialogue content, dialogue content completeness, and dialogue progress information.

[0063] The execution unit is used to output the target response content corresponding to the dialogue request according to the dialogue backtracking strategy.

[0064] According to the above technical means, when the data receiving unit receives a dialogue request, the strategy determination unit determines the dialogue backtracking strategy based on the stored key dialogue information. The device can backtrack the dialogue based on the stored key dialogue information, providing users with the possibility of obtaining the reply content again and improving the user's dialogue experience.

[0065] A third aspect of this application provides a dialogue system, the system comprising:

[0066] Natural Language Understanding Module, Chat Application Module, Large Model Module, Caching and Output Module;

[0067] The natural language understanding module is used to understand the dialogue request and send the understood user intent to the chat application module;

[0068] The chat application module is used to query stored key dialogue information from the cache according to the user's intent, and determine a dialogue backtracking strategy based on the key dialogue information; the key dialogue information includes: dialogue content, dialogue content completeness, and dialogue progress information;

[0069] The chat application module is also used to call the large model module and / or the natural language understanding module according to the dialogue backtracking strategy to determine the target response content corresponding to the dialogue request;

[0070] The output module is used to receive the target reply content sent by the chat application module and output the target reply content.

[0071] Based on the aforementioned technical means, the natural language understanding module in this system receives and identifies dialogue requests, and the chat application module determines a dialogue backtracking strategy based on the stored key dialogue information. It can backtrack the dialogue based on the stored key dialogue information, providing users with the possibility of obtaining the reply content again and improving the user's dialogue experience.

[0072] In some feasible embodiments, the output module includes a voice broadcast module and a graphical interface module;

[0073] If the dialogue backtracking strategy determined by the chat application module is the second dialogue backtracking strategy, then the chat application module requests the natural language understanding module to continue generating semantics. The natural language understanding module identifies the semantics of the previous round of dialogue questions and sends it to the big model module. The big model module generates supplementary response content based on the semantics of the previous round of dialogue questions and sends it to the chat application module. The chat application module sends the supplementary response content to the graphical interface module for display and simultaneously sends the supplementary response content to the voice broadcasting module for broadcasting.

[0074] If the dialogue backtracking strategy determined by the chat application module is the third dialogue backtracking strategy, then the chat application module obtains the first reply content as the target reply content and sends it to the graphical interface module for display, and at the same time sends it to the voice broadcasting module to broadcast the target reply content based on the dialogue progress information.

[0075] If the dialogue backtracking strategy determined by the chat application module is the fourth dialogue backtracking strategy, then the chat application module requests the dialogue question from the previous round from the natural language understanding module. The natural language understanding module identifies the semantics of the dialogue question from the previous round and sends it to the big model module. The big model module generates new response content based on the semantics of the dialogue question from the previous round and sends it to the chat application module. The chat application module sends the new response content to the graphical interface module for display and simultaneously sends the new response content to the voice broadcasting module for broadcasting.

[0076] Based on the aforementioned technical means, different response content can be generated and displayed according to different dialogue backtracking strategies. This allows for targeted output of response content based on the overall situation of the user's previous dialogue, thereby improving the user's dialogue experience.

[0077] A fourth aspect of this application provides an electronic device, comprising:

[0078] The memory is configured to store instructions; and

[0079] The processor is configured to retrieve the instructions from the memory and to implement the dialogue method when executing the instructions.

[0080] A fifth aspect of this application provides a vehicle including a control system that applies the aforementioned dialogue method.

[0081] A sixth aspect of this application provides a machine-readable storage medium storing instructions that cause a machine to perform the aforementioned dialogue method.

[0082] A seventh aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the dialogue method.

[0083] The above technical solution provides a dialogue method that, upon receiving a dialogue request, determines a dialogue backtracking strategy based on stored key dialogue information. This allows for dialogue backtracking based on the stored key dialogue information, providing users with the possibility of obtaining responses again and improving the user's dialogue experience.

[0084] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0085] The accompanying drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the following detailed description to explain the embodiments of this application, but do not constitute a limitation on the embodiments of this application. In the drawings:

[0086] Figure 1 The illustration shows a flowchart of a dialogue method according to an embodiment of this application;

[0087] Figure 2 This illustration schematically shows a diagram illustrating the dialogue method triggering process according to an embodiment of this application;

[0088] Figure 3 This illustration schematically shows a diagram illustrating the execution logic of a dialogue method according to an embodiment of this application;

[0089] Figure 4 A schematic diagram of a dialogue system according to an embodiment of this application is shown.

[0090] Figure 5 This schematically illustrates an interaction diagram of a first backtracking strategy for a dialogue method according to an embodiment of this application.

[0091] Figure 6 This illustration schematically shows an interaction diagram of a second backtracking strategy for a dialogue method according to an embodiment of this application;

[0092] Figure 7 This illustration schematically shows an interaction diagram of a third backtracking strategy for a dialogue method according to an embodiment of this application.

[0093] Figure 8 The diagram illustrates the interaction of the fourth backtracking strategy of the dialogue method according to an embodiment of this application. Detailed Implementation

[0094] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for illustration and explanation of the embodiments of this application and are not intended to limit the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0095] It should be noted that if the embodiments of this application involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0096] Furthermore, if the embodiments of this application involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0097] Terminology Explanation:

[0098] NLU (Natural Language Understanding): Transforming a user's natural language input into a representation that a computer can understand.

[0099] DM (Dialog Manager): Dialogue management is the module responsible for handling the dialogue flow in an intelligent voice interaction system. It parses the user's input and determines the next step in the dialogue process.

[0100] NLG (Natural Language Generation): Natural language generation refers to the ability of intelligent dialogue systems to automatically generate natural language text that conforms to grammatical and semantic rules in response to user input.

[0101] Figure 1The illustration schematically depicts a flowchart of a dialogue method according to an embodiment of this application. This method can be applied to terminal devices such as in-vehicle systems, smartphones, PDAs, tablets, laptops, all-in-one computers, and autonomous driving devices. It is understood that the dialogue method provided in this disclosure can also be applied to other scenarios.

[0102] like Figure 1 As shown in the figure, this application provides a dialogue method, which may include the following steps.

[0103] Step 101: Receive the dialogue request.

[0104] Step 102: Determine the dialogue backtracking strategy based on the dialogue request and the stored key dialogue information; the key dialogue information includes: dialogue content, dialogue content completeness, and dialogue progress information.

[0105] Step 103: Based on the dialogue backtracking strategy, output the target response content corresponding to the dialogue request.

[0106] In some feasible embodiments, a dialogue request generally refers to the user's action of waking up the large dialogue model. The user can wake it up via voice, by triggering a corresponding button, by inputting corresponding text, or even by gestures. If the user wakes up the large model via voice, the dialogue request that triggers dialogue rewind can contain specific words. In some feasible embodiments, the specific word can be "continue," and the dialogue request that triggers dialogue rewind can be a user's voice input such as "continue chatting," "continue broadcasting," "continue the previous topic," or "continue speaking." In other feasible embodiments, the specific word can be "continue," and the dialogue request that triggers dialogue rewind can be a user's voice input such as "continue chatting," "continue broadcasting," "continue the previous topic," or "continue speaking."

[0107] If the user wakes up the large model by pressing a button, then when the user needs to rewind the dialogue, they can trigger the button or icon on the interactive interface that indicates dialogue rewind, or trigger the corresponding button on the device, and the system will send a dialogue request to the large model to trigger dialogue rewind.

[0108] In one feasible embodiment, the large model refers to an in-vehicle AI large model, which is a large-scale artificial intelligence module based on deep learning technology, specifically designed for automotive intelligence. It consists of a massive neural network capable of automatically extracting and learning useful information from massive amounts of data, exhibiting high accuracy, efficiency, and versatility. After training, these modules can process various types of information such as language, images, and speech, and play a crucial role in areas such as autonomous driving and smart cockpits.

[0109] Key dialogue information is information stored during user conversations. In some feasible embodiments, key dialogue information only includes information from one round of dialogue, which includes a dialogue question and its corresponding response. If multiple rounds of dialogue occur during a single conversation until the end, the key dialogue information for the first round is stored, the key dialogue information for the second round is overwritten, and so on, until only the key dialogue information for the last round is stored. When the user starts a second conversation, the key dialogue information from the first round in the second round overwrites the key dialogue information from the last round in the first round. This allows for storing dialogue content for user review with minimal memory usage, reducing system resource consumption. In the aforementioned in-vehicle AI large-scale model, key dialogue information is stored each time the in-vehicle AI large-scale model engages in a conversation. The in-vehicle AI large-scale model's dialogue can be triggered by the user directly calling the voice avatar or by the user operating the in-vehicle host.

[0110] Key dialogue information primarily includes stored dialogue content, dialogue content completeness, dialogue progress information, information displayed on the graphical user interface (GUI), and additional content related to the dialogue question. In some feasible embodiments, the dialogue content includes the dialogue question and the first response. The dialogue question may include: the question raised by the user, the category of the question (car assistant, travel assistant, entertainment assistant, encyclopedia teacher, etc.), etc. The first response is the response from the previous round of dialogue and may include: text on the card (title, number, etc.), video, images, etc. Information displayed on the GUI may include: card type, card size, etc. Additional content related to the dialogue question may include: recommended resources (news, music, movies, apps, etc.), related recommended questions, the source of information on the card, etc. Dialogue progress information may include text output breakpoints, video playback breakpoints, etc. Dialogue content completeness may include indicators indicating that the response content has been generated and indicators indicating that the response content has not been generated, or the proportion of response content generated, etc. This information is cached by the vehicle-mounted system for short-term memory. Key dialogue information is stored in memory as variables. When this information is needed, the corresponding memory address is located based on the variable name, and the value at that memory address is read. Each time the Android application package (APK) restarts, the key dialogue information stored in memory is cleared. This storage method reduces access to the CPU (Central Processing Unit) and hard drive, thus improving response speed. For example, if the dialogue application is a user named "XX" set up on the vehicle's infotainment system, and "XX" closes when the vehicle is powered off and restarts when the vehicle is powered on, the key dialogue information stored in memory will be cleared each time the vehicle is powered on. If the user manually restarts "XX" while the vehicle is continuously powered on, the key dialogue information stored in memory will also be cleared.

[0111] Dialogue backtracking strategy refers to the strategy to be executed when a dialogue request includes at least one of several semantic keywords that continue, extend, supplement, or expand upon the previous dialogue. For example, the received dialogue request might be "Continue chatting," "Add more details," "Expand further," "Anything else?" etc.

[0112] In some feasible embodiments, to inform the user about the conversation rewind feature, the conversation method further includes:

[0113] Upon receiving a signal indicating that the conversation has been interrupted, the interruption signal is simultaneously transmitted to the voice avatar, which then provides a follow-up prompt. For example... Figure 2As shown, if a user triggers the in-vehicle AI large model card during a power-on session and the playback is not completed, a voice conflict interruption scenario is triggered. When the in-vehicle AI large model service receives the interruption signal, it synchronizes the interruption signal to the voice assistant so that the voice assistant can notify the user of the interruption. The interruption signal is canceled after the playback resumes or a new round of question and answer begins. An interruption signal is generated whenever a conversation is interrupted. For example, if the conversation is interrupted while executing a conversation rewind strategy, or if it is interrupted while starting a new conversation, an interruption signal will also be generated. When an interruption signal exists, it is synchronized to the voice assistant. The voice assistant can provide a rewind prompt the next time the user wakes up the voice assistant, or it can provide a rewind prompt the next time a conversation request is received, or it can provide a rewind prompt immediately after receiving an interruption signal.

[0114] If the voice avatar provides a rewind prompt the next time the user wakes it up, or immediately after receiving an interruption signal, the user can choose to abandon the conversation rewind by interacting with the voice avatar. If the user chooses to abandon the conversation rewind through the voice avatar, subsequent prompts for the same interruption signal will not be issued. If the voice avatar provides a rewind prompt the next time a conversation request is received, and the user continues the new conversation, it means the user has given up on understanding the content of the previous conversation. The key information of the subsequent question-and-answer conversation will overwrite the key information stored at the time of the interruption, thus reducing system memory usage. To avoid frequent prompts, each interruption signal will only trigger a rewind prompt once. Therefore, prompting through the voice avatar can help users recall the conversation that existed before the interruption, allowing them to promptly rewind the conversation and obtain the content they want.

[0115] Dialogue recall strategy refers to the different response methods after a dialogue recall is triggered. These response methods are determined based on stored key dialogue information. Target response content refers to the content provided to the user after the dialogue recall strategy is executed; this can be text, video, etc.

[0116] The dialogue content includes the first response. Upon receiving a dialogue request, a dialogue backtracking strategy is determined based on the dialogue request and stored key dialogue information, including:

[0117] 1) Determine if key dialogue information exists to determine if there is dialogue content for backtracking. Key dialogue information is crucial for determining the dialogue backtracking strategy. When the large model on the vehicle's infotainment system is triggered, key dialogue information is stored. If no key dialogue information is stored, it means that no large model dialogue has been triggered before, or the stored key dialogue information has been deleted when the application restarts. In this case, there is no content to backtrack.

[0118] 2) Determine whether the first response content has been fully generated based on the completeness of the dialogue content. Since the large model generates answers in a streaming manner, it takes time for the large model to generate a complete answer. Because the timing of interruption and exit of the large model dialogue is random, the stored key dialogue information will also change due to the content already generated by the large model. In some feasible embodiments, the dialogue content includes a dialogue question and the first response content, which can be text content, i.e., the content displayed on the card. Since the answers generated by the large module are streaming, when the dialogue content completeness in the key dialogue information indicates that the first response content has been fully generated, the vehicle-mounted system receives a complete answer; if the dialogue content completeness in the key information indicates that the first response content has not been fully generated, the vehicle-mounted system does not receive a complete answer. Specifically, the completeness of the first response content can be determined based on the flag corresponding to the dialogue content completeness, or whether the first response content generation rate reaches 100%. In other embodiments, the first response content can also be audio, video, etc. Regardless of the form of the first reply, a completion marker can be added when the large model is generated. The completion marker can be stored in the key information of the dialogue. In this way, the completeness of the first reply can be determined by judging whether the completion marker exists in the key information of the dialogue.

[0119] 3) Based on the dialogue progress information, determine whether the first response content has been completely output. In some feasible embodiments, the output may include voice broadcasting and response content display. Voice broadcasting converts text content into speech and broadcasts it aloud for the user to listen to. Response content display shows the response content on a screen or other display device for the user to view. Generally, the display of response content occurs as the large model is generated, and the chance of a breakpoint in the response content display is relatively small. The speed of voice broadcasting is slower than the speed of response content display, but the content of the voice broadcast corresponds to the text content, so the possibility of a breakpoint in the voice broadcasting is higher. In some feasible embodiments, the position of the voice broadcasting breakpoint is used to determine whether the first response content has been completely output. Generally, if the voice broadcasting breakpoint is at the end of the text content, it can be determined that the first response content has been completely output; otherwise, the first response content has not been completely output.

[0120] 4) Determine the corresponding dialogue backtracking strategy based on whether there is dialogue content for backtracking, whether the first response content has been fully generated, and whether the first response content has been fully output, specifically including:

[0121] If no dialogue content is available for backtracking, then determine and execute the first dialogue backtracking strategy;

[0122] If there is dialogue content for backtracking, and the first response content has been generated and output, then the second dialogue backtracking strategy is determined and executed.

[0123] If there is dialogue content for backtracking, and the first response content has been generated but not yet output, then the third dialogue backtracking strategy is determined and executed.

[0124] If there is dialogue content for backtracking, and the first response content has not been fully generated, then the fourth dialogue backtracking strategy is determined and executed.

[0125] Based on the aforementioned technical means, the state of the previous round of dialogue obtained by analyzing key information in the dialogue can serve as the basis for determining the dialogue backtracking strategy. This identifies the user's actual need to trigger dialogue backtracking, and different dialogue backtracking strategies are executed according to different needs to maximize the satisfaction of the user's dialogue backtracking needs and improve the user's dialogue experience.

[0126] In some feasible embodiments, the first dialogue backtracking strategy may include: starting a new dialogue. Even if there is no dialogue content for backtracking (i.e., no previous dialogue), but the user has issued a dialogue request, the large model cannot ignore it. Therefore, when there is no dialogue content for backtracking, the large model can start a new dialogue with the user to respond to the user's dialogue needs, ensuring that every request is responded to. The response content generated by this new dialogue is the target response content corresponding to the dialogue request in the first dialogue backtracking strategy.

[0127] The second dialogue backtracking strategy may include: obtaining the dialogue questions and the first response from the key dialogue information;

[0128] The dialogue question and the first response content are synchronized in context;

[0129] Based on the synchronized dialogue question and the first response, supplementary response content is generated. This supplementary response content is the target response content corresponding to the dialogue request in the second dialogue backtracking strategy. In the second dialogue backtracking strategy, the target response content may be partially the same as the first response content, or it may be completely different.

[0130] There is dialogue content available for backtracking, meaning there is a previous round of dialogue. Once the dialogue content has been generated and the first response has been output, it is determined that the user has consumed all the content from the previous round of dialogue. Therefore, the user's purpose in triggering dialogue backtracking is semantically "to continue generating," meaning the user wants to obtain more content based on what has already been displayed. At this point, the large model supplements the previous response with more content, providing the user with more of the content they need and improving the user experience.

[0131] The third dialogue backtracking strategy may include: obtaining the first response content and dialogue progress information from the key dialogue information, wherein the first response content is used as the target response content in the third dialogue backtracking strategy.

[0132] If there is dialogue content for backtracking, and the first response content has been generated but not yet fully output, in this case, the large model has already generated complete response content, but the user has not finished listening to the voice broadcast or reading the card content. At this time, the user's intention to trigger dialogue backtracking should correspond to the semantics of "re-displaying and continuing the broadcast." Therefore, it is necessary to obtain the first response content and dialogue progress information from the key information of the dialogue to prepare content for the output of subsequent strategy execution results, so that the user can continue to read or listen to the answer that was read or heard halfway in the previous round of dialogue, thereby improving the user experience.

[0133] The fourth dialogue backtracking strategy may include:

[0134] The dialogue question from the previous round is retrieved from the key information of the dialogue. A new response is generated based on this question, and this new response is the target response content corresponding to the dialogue request in the fourth dialogue backtracking strategy. In the fourth dialogue backtracking strategy, the target response content may be partially the same as the first response content, or it may be completely different.

[0135] Since the answers generated by the large model are streamed, if the first response in the key information is incomplete, the vehicle's system will not receive a complete answer. Since the first response is not fully generated, the large model can only answer the same question again from the previous round and then broadcast normally. Although the consistency of the content generated by the two responses cannot be guaranteed, it can provide a complete answer to the dialogue questions that the user did not receive before, thus improving the user experience.

[0136] In some feasible embodiments, the step of outputting the target response content corresponding to the dialogue request according to the dialogue backtracking strategy includes:

[0137] The dialogue backtracking strategy is executed to obtain the target response content;

[0138] Based on the target response content, output the target response content according to the output strategy;

[0139] The output strategy includes a voice broadcast strategy and a response content display strategy. The voice broadcast strategy primarily involves broadcasting the target response content after the execution of the guidance information and the dialogue rewind strategy. The guidance information is a pre-configured dialogue template; regardless of when dialogue rewind is performed, the same dialogue rewind strategy will semantically broadcast the same dialogue template. The response content after the execution of the dialogue rewind strategy varies depending on the specific rewind strategy used.

[0140] If the first dialogue backtracking strategy is used in the dialogue, then when implementing the voice broadcast strategy, the response content of the new dialogue needs to be broadcast aloud. If the dialogue question is what varieties of roses are there, then the answer about rose varieties will be broadcast. If the second dialogue backtracking strategy is used in the dialogue, then when implementing the voice broadcast strategy, the supplementary content of the response needs to be broadcast aloud. If the dialogue question is what common diseases of roses are there, and the previous round of dialogue answered powdery mildew and rust, and the supplementary content answered brown spot and black spot, then when implementing the voice broadcast strategy, the answer including brown spot and black spot needs to be broadcast. If the third dialogue backtracking strategy is used in the dialogue, then when implementing the voice broadcast strategy, the target response content needs to be broadcast aloud based on the dialogue progress information. If the dialogue question is what common diseases of roses are there, and the previous round of dialogue answered powdery mildew and rust, and the supplementary content answered brown spot and black spot, then the answer including brown spot and black spot needs to be broadcast. If the third dialogue backtracking strategy is used in the dialogue, then when implementing the voice broadcast strategy, the target response content needs to be broadcast aloud based on the dialogue progress information. If the response in the dialogue round is about powdery mildew and rust, but the dialogue output only shows half of the response related to powdery mildew, then when executing the voice broadcast strategy, it is necessary to continue broadcasting the other half of the response related to powdery mildew and the response related to rust, according to the dialogue progress information. If the fourth dialogue backtracking strategy is executed in the dialogue, then when executing the voice broadcast strategy, it is necessary to broadcast the new response. If the dialogue question is about common diseases of roses, and the new response mainly introduces powdery mildew and brown spot, then when executing the voice broadcast strategy, it is necessary to broadcast the response that introduces powdery mildew and brown spot.

[0141] The response content display strategy primarily showcases the execution results of the dialogue backtracking strategy. This can be a simple text display of the target response content or a gesture-based image display. If it's a text display, it can be categorized into new dialogue display and re-display display based on whether the response content has already been shown to the user in the dialogue backtracking strategy. The new dialogue display and re-display display methods can be customized according to requirements. For example, the new dialogue display method can have a special style, such as a typewriter effect that gradually displays the response content as it is generated. The re-display method can be without a special style, such as directly displaying all the content. Directly displaying all the content allows users to directly view all the response content. If there is voice prompts, users can quickly locate the output breakpoint in the response content, avoiding reading delays caused by other effects. If there are no voice prompts, users can quickly browse the content they have already read and then carefully study the unread content later.

[0142] In some feasible embodiments, the output strategy may include only a voice broadcast strategy or only a response content display strategy. The response content display strategy can be divided into text display strategy, image display strategy, video display strategy, etc., according to different presentation formats. The presentation of the same content can be a single form or a combination of two or more forms, such as single text display, single image display, single video display, or a combination of text and image display, text and video display, image and video display, or a combination of text, image, and video display. Therefore, the output strategy can also be a combination of a voice broadcast strategy and any response content display strategy, such as a combination of a voice broadcast strategy and a text display strategy; or a combination of a voice broadcast strategy, a text display strategy, and an image display strategy.

[0143] In some feasible embodiments, the voice broadcasting strategy includes guiding information output and target response content output, which are linked by listening for a guiding information output end signal. Executing the target response content output only after the guiding information output end signal is detected avoids interruption of the guiding information output, thus improving the completeness of the user's listening experience. In one feasible embodiment, a voice broadcasting end function (ttsend) can be called to issue an end signal after the guiding information output ends. Detecting the ttsend call confirms that the previous voice broadcast has ended.

[0144] The guidance information may include a new conversation start prompt voice when the first conversation backtracking strategy is executed, and a continuation prompt voice for the original conversation when the third or fourth conversation backtracking strategy is executed. The guidance information is a pre-configured conversation template that serves to connect conversations and guide the user in the direction of the next part of the conversation. For example, a new conversation start prompt voice could be something like, "Okay, you start a topic," or "Okay, what do you want to talk about?" After receiving such a response, the user will naturally say what they want to talk about. The big model then generates a response based on this content to start a new conversation. Similarly, a continuation prompt voice could be something like, "Okay, let's continue with the previous topic." After receiving such a response, the user knows they will hear the answer to the topic they previously asked.

[0145] The target response content output may include: the output of the response content for the new dialogue after executing the first dialogue backtracking strategy, specifically including the voice broadcast of the response content of the new dialogue.

[0146] Alternatively, the output of the supplementary response content after executing the second dialogue backtracking strategy may specifically include the voice broadcast of the supplementary response content.

[0147] Alternatively, after executing the third dialogue backtracking strategy, the output of the first response content can continue based on the dialogue progress information. Specifically, this may include continuing the voice playback of the response content based on the dialogue output breakpoint position. In some feasible embodiments, continuing the voice playback of the response content based on the dialogue output breakpoint position can be achieved by first determining the sentence where the dialogue output breakpoint is located based on the dialogue output breakpoint position, and then performing voice playback from the sentence where the output breakpoint is located.

[0148] Alternatively, the output of the new response content after executing the fourth dialogue backtracking strategy may specifically include the voice broadcast of the new response content.

[0149] The response content display strategy mainly focuses on displaying the execution results, specifically including:

[0150] After executing the first dialogue backtracking strategy, the response content of the new dialogue is displayed according to the display method of the new dialogue. In one feasible embodiment, the new dialogue is displayed gradually using a typewriter effect.

[0151] After executing the second dialogue rewind strategy, the supplementary response content is displayed in the manner of the new dialogue. In one feasible embodiment, the new dialogue is displayed gradually using a typewriter effect.

[0152] Alternatively, after executing the third dialogue backtracking strategy, the first response content can be displayed again. In some feasible embodiments, the display again method is to directly display all response content.

[0153] Alternatively, after executing the fourth dialogue backtracking strategy, the new response content can be displayed according to the display method of the new dialogue. In one feasible embodiment, the new dialogue is displayed gradually using a typewriter effect.

[0154] In some feasible embodiments, the key dialogue information is stored in the following manner:

[0155] When generating response content at the start of the current dialogue round, the key dialogue information from the previous round is cleared from the cache, and the key dialogue information for the current round is stored in the cache. This current dialogue round can be a backtracking dialogue or any other dialogue without backtracking. For any dialogue, the key dialogue information is cleared when generating the corresponding response content. If a dialogue backtracks, the dialogue questions and responses from the previous round have already been obtained from the key dialogue information; clearing the key dialogue information will not affect the backtracking and will also ensure that the key dialogue information for the current round is accurately stored.

[0156] like Figure 3As shown, upon receiving a dialogue request that triggers dialogue rewind, the system first checks if there is any dialogue content for rewind in the storage. If there is no dialogue content for rewind, it directly plays a new dialogue start prompt to obtain the basis for generating the response content. If there is dialogue content for rewind, it further checks if the first response content is complete to determine if the first response content has been generated. If the first response content has not been generated, it restarts the generation of the response content and replays it. During this process, there are GPT generation in progress tags, and the facial expressions of the voice playback content are inherited. If the first response content has been generated, it further checks if the output of the first response content has been completed. If the first response content has been output, it re-requests semantics and continues to generate. During this process, there are GPT generation in progress tags, and the facial expressions of the voice playback content are inherited. If the first response content has not been output, the card is reopened, and output continues from the breakpoint of the first response content output. During this process, there are GPT generation in progress tags, and the facial expressions of the previous voice playback content are inherited.

[0157] This application embodiment also provides another dialogue method, the method including:

[0158] Step 201: Receive the dialogue request;

[0159] Step 202: Determine the dialogue backtracking strategy based on the dialogue request and the stored key dialogue information; the key dialogue information includes: dialogue content, dialogue content completeness, and dialogue progress information;

[0160] Step 203: Based on the dialogue backtracking strategy, output the target response content corresponding to the dialogue request.

[0161] In this embodiment, the dialogue backtracking strategy includes: a first dialogue backtracking strategy, a second dialogue backtracking strategy, a third dialogue backtracking strategy, and a fourth dialogue backtracking strategy. During the execution of the dialogue backtracking strategy, key dialogue information needs to be stored. In some feasible embodiments, when the dialogue backtracking strategy is the first dialogue backtracking strategy, the key dialogue information of the new dialogue is stored in a cache. For example, upon receiving "Continue chatting," it is determined that there is no dialogue content for backtracking, and then the guiding information "Okay, you can start a topic" is played. Before playing the guiding information, the dialogue request "Continue chatting" needs to be stored in the cache, and the response guiding information "Okay, you can start a topic" needs to be stored in the cache. When the dialogue question "What are some common diseases of roses?" is received, when generating the response content, the stored "Continue chatting" and "Okay, you can start a topic" are cleared, and "What are some common diseases of roses?" is stored in the cache. Subsequently, the stored key dialogue information is gradually improved according to the progress of the response content generation.

[0162] When the dialogue backtracking strategy is the second dialogue backtracking strategy, upon starting to generate supplementary response content, the key dialogue information of the previous round of dialogue is cleared from the cache, and the dialogue request and the generated supplementary response content are stored in the cache. For example, if the dialogue question of the previous round of dialogue was "What are the common diseases of roses?", and the response content of the previous round of dialogue answered powdery mildew and rust, upon receiving "Continue chatting", the dialogue backtracking strategy determines that a supplementary answer to "What are the common diseases of roses?" is needed. Therefore, the stored dialogue question and response content are cleared, and then "Continue chatting" and the generated supplementary content are stored in the cache.

[0163] When the dialogue backtracking strategy is the third dialogue backtracking strategy, after obtaining the first response content from the key dialogue information, the key dialogue information of the previous round of dialogue is cleared from the cache, the dialogue request is stored in the cache, and the first response content is stored in the cache as the target response content. For example, if the dialogue question of the previous round of dialogue is "What are the common diseases of roses?", and the response content of the previous round of dialogue is "powdery mildew and rust", but the dialogue output only outputs half of the response content related to powdery mildew, after receiving "Continue chatting", according to the dialogue backtracking strategy, it is determined that it is necessary to continue to answer "What are the common diseases of roses?", then after obtaining the response content of the previous round, the dialogue question and response content stored in the cache are cleared, and then "Continue chatting" and the response content of the previous round are stored in the cache as the target response content, that is, the response content generated in the previous round that includes powdery mildew and rust is stored.

[0164] When the dialogue backtracking strategy is the fourth dialogue backtracking strategy, when generating new response content, the key dialogue information of the previous round of dialogue is cleared from the cache, and the dialogue request and the generated new response content are stored in the cache. If the dialogue question of the previous round of dialogue was "What are the common diseases of roses?", and the response content of the previous round of dialogue was "powdery mildew and rust", but the response content was not fully generated, upon receiving "Continue chatting", the dialogue backtracking strategy determines that an answer to "What are the common diseases of roses?" is required. Therefore, when generating new response content, the key dialogue information of the previous round of dialogue is cleared from the cache, and "Continue chatting" and the generated new response content are stored in the cache.

[0165] Based on the above technical means, different dialogue backtracking strategies correspond to different storage methods. When a dialogue backtracking strategy needs to be generated, the key dialogue information of the previous round of dialogue is cleared from the cache when the generation begins, and the dialogue request and the generated content are stored in the cache to ensure that the key dialogue information is updated in real time. When the dialogue backtracking strategy needs to backtrack and output the response content of the previous round, the first response content is stored in the cache as the target response content to ensure that the key dialogue information can be stored normally, and also to ensure the accuracy of the key dialogue information.

[0166] This application embodiment also provides another dialogue method, the dialogue method including:

[0167] Step 301: Receive the dialogue request.

[0168] Step 302: Identify the intent of the dialogue request. The intent of the dialogue request refers to the purpose that the dialogue request hopes to achieve. In this application, the intent of the dialogue request refers to the function that the requested device hopes to perform. For example, when the requested device is an in-vehicle system, the intent of the dialogue request can be the control of a functional unit on the in-vehicle system, or it can be a casual conversation.

[0169] Step 303: Determine whether the stated intent is within the target vertical domain. A vertical domain refers to a specific area categorized based on professional skills and business characteristics. Taking in-vehicle systems as an example, it can be divided into casual conversation vertical domains, entertainment vertical domains, control vertical domains, etc. In this embodiment, the target vertical domain refers to the casual conversation vertical domain, i.e., the vertical domain where the conversation takes place.

[0170] Step 304: When the intent is the intent of the target vertical domain, determine whether the dialogue request triggers dialogue backtracking.

[0171] When the intent of a dialogue request is to engage in casual conversation, the request can be either an intention to directly conduct a question-and-answer session or an intention to revisit the previous conversation. For the intention to directly conduct a question-and-answer session, a response needs to be generated based on the received dialogue request and displayed to the user. In some feasible embodiments, dialogue revisiting is triggered when the dialogue request includes at least one of several semantic keywords that continue, extend, supplement, or expand upon the previous conversation.

[0172] Step 305: When triggering dialogue backtracking, determine the dialogue backtracking strategy based on the dialogue request;

[0173] Step 306: Based on the dialogue backtracking strategy, output the target response content corresponding to the dialogue request.

[0174] Upon receiving a dialogue request, the system can determine whether to trigger dialogue backtracking based on the identified intent. If dialogue backtracking is not triggered, the system can respond directly. If dialogue backtracking is triggered, the system can execute steps to determine the dialogue backtracking strategy. This avoids consuming system memory by executing steps to determine the dialogue backtracking strategy when the dialogue request does not trigger dialogue backtracking.

[0175] In some feasible embodiments, when a dialogue backtracking is triggered, the dialogue backtracking strategy can be determined based on the dialogue request and the stored key dialogue information. The specific methods will not be described in detail here.

[0176] A second aspect of this application provides a dialogue device, the dialogue device comprising:

[0177] The data receiving unit is used to receive dialogue requests;

[0178] The strategy determination unit is used to determine a dialogue backtracking strategy based on the dialogue request and stored key dialogue information; the key dialogue information includes: dialogue content, dialogue content completeness, and dialogue progress information.

[0179] The execution unit is used to output the target response content corresponding to the dialogue request according to the dialogue backtracking strategy.

[0180] According to the above technical means, when the data receiving unit receives a dialogue request, the strategy determination unit determines the dialogue backtracking strategy based on the stored key dialogue information. The device can backtrack the dialogue based on the stored key dialogue information, providing users with the possibility of obtaining the reply content again and improving the user's dialogue experience.

[0181] In some feasible embodiments, the dialogue content includes first response content; the strategy determination unit includes a judgment module, which is used to:

[0182] Determine if there is key information in the dialogue to determine if there is dialogue content that can be used for backtracking;

[0183] Based on the completeness of the dialogue content, determine whether the first response content has been fully generated;

[0184] Based on the dialogue progress information, determine whether the first reply content has been completely output.

[0185] In some feasible embodiments, the dialogue backtracking unit includes a policy determination module, used to determine a dialogue backtracking policy based on the judgment result of the judgment module, specifically:

[0186] If no dialogue content is available for backtracking, then determine and execute the first dialogue backtracking strategy;

[0187] If there is dialogue content for backtracking, and the first response content has been generated and output, then the second dialogue backtracking strategy is determined and executed.

[0188] If there is dialogue content for backtracking, and the first response content has been generated but not yet output, then the third dialogue backtracking strategy is determined and executed.

[0189] If there is dialogue content for backtracking, and the first response content has not been fully generated, then the fourth dialogue backtracking strategy is determined and executed.

[0190] In some feasible embodiments,

[0191] The first dialogue backtracking strategy includes: starting a new dialogue;

[0192] The second dialogue backtracking strategy includes: obtaining the previous dialogue question and the first response from the key dialogue information;

[0193] Generate supplementary response content based on the dialogue question and the first response content;

[0194] The third dialogue backtracking strategy includes: obtaining the first response content and dialogue progress information from the key dialogue information;

[0195] The fourth dialogue backtracking strategy includes:

[0196] The dialogue questions from the previous round are obtained from the key information of the dialogue, and new response content is generated based on the dialogue questions.

[0197] In some feasible embodiments, the execution unit further includes: a response content generation module and an output display module;

[0198] The response content generation module is used to execute the dialogue backtracking strategy to obtain the target response content;

[0199] The output display module is used to output the target response content based on the output strategy.

[0200] The output strategy includes a voice broadcast strategy and a response content display strategy.

[0201] The output display module includes a voice broadcast module and an interface display module, and the output strategy includes a voice broadcast strategy and a response content display strategy.

[0202] In some feasible embodiments, the voice broadcasting strategy includes guiding information output and target response content output, wherein the guiding information output and the target response content output are linked by listening to the end signal of the guiding information output.

[0203] The guidance information includes a new dialogue start prompt when the first dialogue backtracking strategy is executed, and a continuation prompt when the original dialogue is executed when the third or fourth dialogue backtracking strategy is executed. The guidance information is a pre-configured dialogue template that serves to connect dialogues and guide the user in the direction of the next part of the conversation.

[0204] The target response content output includes: the output of the response content for the new dialogue after executing the first dialogue backtracking strategy, which may specifically include the voice broadcasting module broadcasting the response content for the new dialogue.

[0205] Alternatively, after executing the second dialogue backtracking strategy, the output of the supplementary response content may specifically include the voice broadcast module broadcasting the supplementary response content via voice.

[0206] Alternatively, after executing the third dialogue backtracking strategy, the output of the first response content can continue based on the dialogue progress information. Specifically, this may include the voice broadcasting module continuing to broadcast the response content based on the dialogue progress information.

[0207] Alternatively, after executing the fourth dialogue backtracking strategy, the output of the new response content may specifically include a voice broadcast module broadcasting the new response content.

[0208] The response content display strategy is executed through the interface display module, including:

[0209] After executing the first dialogue backtracking strategy, the interface display module will display the response content of the new dialogue in the same way as the new dialogue.

[0210] After executing the second dialogue backtracking strategy, the interface display module will display the supplementary content in the manner of displaying the new dialogue.

[0211] Alternatively, after executing the third dialogue backtracking strategy, the interface display module will display the first reply content in the manner of re-displaying it.

[0212] Alternatively, after executing the fourth dialogue backtracking strategy, the interface display module will display the new response content according to the display method of the new dialogue.

[0213] In some feasible embodiments, the execution unit further includes a storage unit, which is used to clear the key dialogue information of the previous round of dialogue in the cache and store the key dialogue information of the current round of dialogue in the cache when the reply content is generated at the beginning of the current round of dialogue.

[0214] In some feasible embodiments, the storage unit stores key dialogue information according to the execution of different dialogue backtracking strategies. When the dialogue backtracking strategy is the first dialogue backtracking strategy, the storage unit stores the key dialogue information of the new dialogue in the cache.

[0215] When the dialogue backtracking strategy is the second dialogue backtracking strategy, when the supplementary response content is generated, the storage unit clears the key dialogue information of the previous round of dialogue in the cache and stores the dialogue request and the generated supplementary response content in the cache.

[0216] When the dialogue backtracking strategy is the third dialogue backtracking strategy, after obtaining the first reply content from the dialogue key information, the storage unit clears the dialogue key information of the previous round of dialogue in the cache, stores the dialogue request in the cache, and stores the first reply content as the target reply content in the cache.

[0217] When the dialogue backtracking strategy is the fourth dialogue backtracking strategy, when generating new response content, the storage unit clears the key dialogue information of the previous round of dialogue in the cache and stores the dialogue request and the generated new response content in the cache.

[0218] A third aspect of this application provides a dialogue system, such as Figure 4 As shown, the system includes: a natural language understanding module, a chat application module, a large model module, a cache, and an output module;

[0219] The Natural Language Understanding (NLU) module is used to understand dialogue requests and send the understood user intent to the chat application module.

[0220] The chat application module ChatCopilot is used to query stored key dialogue information from the cache according to the user's intent, and determine a dialogue backtracking strategy based on the key dialogue information; the key dialogue information includes: dialogue content, dialogue content completeness, and dialogue progress information;

[0221] The chat application module is also used to call the large model module and / or the natural language understanding module according to the dialogue backtracking strategy to determine the target response content corresponding to the dialogue request;

[0222] The output module is used to receive the target reply content sent by the chat application module and output the target reply content.

[0223] The natural language understanding module supports machine understanding of text content. In the application embodiment, the natural language understanding module implements semantic recognition of received dialogue requests or dialogue questions. The chat application module provides an interface for the vehicle-side to call the large model module.

[0224] In some feasible embodiments, the system further includes an interaction module and a dialogue management module (DM). During the dialogue, the interaction module is used to enable the interaction between the natural language understanding module and the large model module and the dialogue management module. The dialogue management module is used to fully wake up the chat application module. The user intent understood by the natural language understanding module is first sent to the interaction module, and then the interaction module sends the user intent to the backend dialogue management module. The dialogue management module then sends the user intent to the chat application module.

[0225] In some feasible embodiments, the dialogue content includes a first reply; the chat application module determines the dialogue backtracking strategy in the following manner:

[0226] Determine if there is key information in the dialogue to determine if there is dialogue content that can be used for backtracking;

[0227] Based on the completeness of the dialogue content, determine whether the first response content has been fully generated;

[0228] Based on the dialogue progress information, determine whether the first response content has been completely output;

[0229] The dialogue backtracking strategy is determined based on whether there is dialogue content for backtracking, whether the first response content has been generated, and whether the first response content has been output.

[0230] If no dialogue content is available for backtracking, then determine and execute the first dialogue backtracking strategy;

[0231] If there is dialogue content for backtracking, and the first response content has been generated and output, then the second dialogue backtracking strategy is determined and executed.

[0232] If there is dialogue content for backtracking, and the first response content has been generated but not yet output, then the third dialogue backtracking strategy is determined and executed.

[0233] If there is dialogue content for backtracking, and the first response content has not been fully generated, then the fourth dialogue backtracking strategy is determined and executed.

[0234] In some feasible embodiments, the chat application module is further configured to store key dialogue information in a cache after receiving the dialogue intent identified by the natural language understanding module. When the chat application module receives the dialogue intent identified by the natural language understanding module, it indicates that it is about to start generating response content. If dialogue rewinding is required, the dialogue questions and responses from the previous round of dialogue have already been obtained from the key dialogue information. Clearing the key dialogue information will not affect the dialogue rewinding, and it also ensures that the key dialogue information of the current round of dialogue can be accurately stored. Specifically:

[0235] After determining the dialogue backtracking strategy, when the dialogue backtracking strategy determined by the chat application module is the first dialogue backtracking strategy, during the process of the chat application module calling the large model module, natural language understanding module and output module to conduct a new dialogue, the chat application module stores the key dialogue information of the new dialogue in the cache.

[0236] When the chat application module determines the dialogue backtracking strategy as the second dialogue backtracking strategy, when the large model module starts generating supplementary reply content, the chat application module clears the key dialogue information of the previous round of dialogue in the cache and stores the dialogue request and the generated supplementary reply content in the cache.

[0237] When the conversation backtracking strategy determined by the chat application module is the third conversation backtracking strategy, after the chat application module retrieves the first reply content from the cache, the chat application module clears the key conversation information of the previous round of conversation in the cache, stores the conversation request in the cache, and stores the first reply content as the target reply content in the cache.

[0238] When the chat application module determines the dialogue backtracking strategy as the fourth dialogue backtracking strategy, when the large model module starts generating new reply content, the chat application module clears the key dialogue information of the previous round of dialogue in the cache, and stores the dialogue request and the generated new reply content in the cache.

[0239] Based on the above technical means, different dialogue backtracking strategies correspond to different storage methods. When a dialogue backtracking strategy needs to be generated, the key dialogue information of the previous round of dialogue is cleared from the cache when the generation begins, and the dialogue request and the generated content are stored in the cache to ensure that the key dialogue information is updated in real time. When the dialogue backtracking strategy needs to backtrack and output the response content of the previous round, the first response content is stored in the cache as the target response content to ensure that the key dialogue information can be stored normally, and also to ensure the accuracy of the key dialogue information.

[0240] In some feasible embodiments, the output module includes a text-to-speech (TTS) module and a graphical user interface (GUI) module.

[0241] If the dialogue backtracking strategy determined by the chat application module is the first dialogue backtracking strategy, the natural language understanding module identifies the semantics of the new dialogue question and sends it to the big model module. The big model module generates new response content based on the semantics of the new dialogue question and sends it to the chat application module. The chat application module sends the new response content to the graphical interface module for display and simultaneously sends the new response content to the voice broadcasting module for broadcasting.

[0242] If the dialogue backtracking strategy determined by the chat application module is the second dialogue backtracking strategy, then the chat application module requests the natural language understanding module to continue generating semantics. The natural language understanding module identifies the semantics of the previous round of dialogue questions and sends it to the big model module. The big model module generates supplementary response content based on the semantics of the previous round of dialogue questions and sends it to the chat application module. The chat application module sends the supplementary response content to the graphical interface module for display and simultaneously sends the supplementary response content to the voice broadcasting module for broadcasting.

[0243] If the dialogue backtracking strategy determined by the chat application module is the third dialogue backtracking strategy, then the chat application module obtains the first reply content as the target reply content and sends it to the graphical interface module for display, and at the same time sends it to the voice broadcasting module to broadcast the target reply content based on the dialogue progress information.

[0244] If the dialogue backtracking strategy determined by the chat application module is the fourth dialogue backtracking strategy, then the chat application module requests the dialogue question from the previous round from the natural language understanding module. The natural language understanding module identifies the semantics of the dialogue question from the previous round and sends it to the big model module. The big model module generates new response content based on the semantics of the dialogue question from the previous round and sends it to the chat application module. The chat application module sends the new response content to the graphical interface module for display and simultaneously sends the new response content to the voice broadcasting module for broadcasting.

[0245] Based on the aforementioned technical means, different response content can be generated and displayed according to different dialogue backtracking strategies. This allows for targeted output of response content based on the overall situation of the user's previous dialogue, thereby improving the user's dialogue experience.

[0246] In some feasible embodiments, the system further includes a Natural Language Generation (NLG) module. The NLG module is used to broadcast guiding information, and the broadcasting of guiding information and the broadcasting of the target response content are linked by listening for a signal indicating the end of the guiding information broadcast.

[0247] The guidance information broadcast includes a new dialogue start prompt voice broadcast when the natural language generation module executes the first dialogue backtracking strategy, and an original dialogue continuation prompt voice broadcast when the third or fourth dialogue backtracking strategy is executed.

[0248] In some feasible embodiments, both the natural language understanding module and the large model module are deployed in the cloud, and the natural language understanding module and the large model module interact with the chat application module through the interaction module and the dialogue management module.

[0249] The following section uses the text content of the reply as an example to further explain the specific interaction process of the dialogue method provided in this application in conjunction with the dialogue system.

[0250] If no dialogue content exists for backtracking, determine and execute the first dialogue backtracking strategy. For example... Figure 5 The diagram shown is the interaction diagram of the first backtracking strategy in this application. Figure 5 As shown, the user triggers the Natural Language Understanding (NLU) module to recognize the dialogue rewind intention by using the voice command "Continue Chat" containing the trigger word "continue." The NLU module then sends the dialogue rewind intention to the interaction module, which in turn wakes up the Dialogue Management (DM) module and sends the rewind intention back to it. The DM module then fully wakes up the Chat Application (ChatCopilot) module and sends the rewind intention back. Upon receiving the rewind intention, the Chat Application module checks its cache for rewind content and determines if dialogue rewind is supported. If no rewind content is found, the Chat Application module returns a success code (successful_no_content) to the DM module, which in turn returns a success code (successful_no_content) to the Natural Language Generation (NLG) module. The NLG module then announces the start of a new conversation with the voice command "Okay, you can start a topic." When no rewind content is found, it can be determined that there is no previous conversation. In this case, playing the new conversation start announcement allows the system to obtain the content the user wants to discuss, thus enabling it to tailor the conversation to the user's needs and improve the user experience.

[0251] If there is dialogue content for backtracking, and the first response content has been generated and output, meaning there is backtracking content, the user has seen all the content and listened to the voice broadcast, then it is assumed that the user has already consumed the current question and answer. Therefore, the "continue chatting" intention should correspond to the semantics of "continue generating". So we make a second semantic request, which must send the semantics of "continue generating", and then determine and execute the second dialogue backtracking strategy.

[0252] like Figure 6 The diagram shown is the interaction diagram of the second backtracking strategy in this application. Figure 6As shown, the user triggers the natural language understanding module to recognize the dialogue rewind intention by using the voice command "Continue Chat" containing the trigger word "continue," and sends the dialogue rewind intention to the interaction module by issuing a search content command. The interaction module wakes up the dialogue management module and sends the dialogue rewind intention to the dialogue management module. The dialogue management module fully wakes up the chat application module and sends the dialogue rewind intention. Upon receiving the dialogue rewind intention, the chat application module starts to check if there is rewind content in the cache, determines whether dialogue rewind is supported, and if there is rewind content in the cache, the content is generated, and the voice playback is completed, the chat application module returns a success code (successful_tts_finished) to the dialogue management module. Simultaneously, the chat application module sends a request to the natural language understanding module to continue generating semantics. This request is synchronized to the natural language understanding module via the interaction module. The natural language understanding module identifies the dialogue question and issues a search command, synchronizing the understanding results of the previous dialogue question and the first response to the chat application module via the interaction module and the dialogue management module. It is also sent to the big model module. The big model module generates supplementary response content based on the understanding results and sends it to the chat application module via the interaction module and the dialogue management module. The chat application module provides the supplementary response content to the graphical user interface (GUI) module, which gradually displays the cards using a typewriter effect. It also provides it to the voice broadcast module, which broadcasts the supplementary response content.

[0253] If there is dialogue content for backtracking, and the first response content has been generated but not yet output, the end side has already received the completed answer, but the user has not finished listening to the voice broadcast or reading the card content. In this case, the "continue chatting" intent should correspond to the semantics of "re-display and continue broadcasting", and the third dialogue backtracking strategy is determined and executed.

[0254] like Figure 7 The diagram shown is the interaction diagram of the third backtracking strategy in this application. Figure 7As shown, the user triggers the natural language understanding module to recognize the dialogue reversal intent by using the voice command "Continue chatting" containing the trigger word "continue," and sends the dialogue reversal intent to the interaction module by issuing a search content command. The interaction module wakes up the dialogue management module and sends the dialogue reversal intent to the dialogue management module. The dialogue management module fully wakes up the chat application module and sends the dialogue reversal intent. Upon receiving the dialogue reversal intent, the chat application module checks if there is reversal content in the cache and determines whether dialogue reversal is supported. If there is reversal content in the cache, the content is generated, but the voice playback is not complete, the chat application module returns a success code (successful) to the dialogue management module. The dialogue management module returns a success code (successful) to the natural language generation module, which then plays the original dialogue continuation prompt voice. For example, after playing "Okay, let's continue talking about the previous topic," it notifies the dialogue management module, which then notifies the chat application module. The chat application module provides the previous round of responses to the graphical interface module for direct display of cards. At the same time, the chat application module provides the previous round of responses and the dialogue output breakpoint position to the voice playback module, which plays the content after the interruption.

[0255] If there is dialogue content for backtracking, and the first response content has not been fully generated, the end side has not received a complete answer because the answer generated by the large model module is streaming. Therefore, the large model module can only answer the same question again in the previous round. At this time, the fourth dialogue backtracking strategy is determined and executed.

[0256] like Figure 8 The diagram shown is the interaction diagram of the fourth backtracking strategy in this application. Figure 8As shown, the user triggers the vehicle's natural language understanding module to recognize the dialogue reversal intent by using the voice command "Continue Chat" containing the trigger word "continue." The module then sends the dialogue reversal intent to the interaction module, which in turn wakes up the dialogue management module and sends the reversal intent back to it. The dialogue management module then fully wakes up the chat application module and sends the reversal intent back to it. Upon receiving the reversal intent, the chat application module checks if there is any reversal content in the cache and determines whether dialogue reversal is supported. If there is reversal content in the cache but it has not been fully generated and the voice playback is incomplete, the chat application module returns a success code (successful) to the dialogue management module. The dialogue management module then returns a success code (successful) to the natural language generation module, which then plays the original dialogue continuation prompt. For example, after the announcement "Okay, let's continue our previous topic" is completed, the dialogue management module is notified. The dialogue management module then notifies the chat application module. The chat application module requests the dialogue question (query) from the previous round of question-and-answer, which is synchronized to the natural language understanding module via the interaction module. The natural language understanding module recognizes the dialogue question and issues a search command. The understanding result is synchronized to the chat application module via the interaction module and the dialogue management module, and simultaneously sent to the big model module. The big model module generates a response based on the understanding result and sends it to the chat application module via the interaction module and the dialogue management module. The chat application module provides the new response to the graphical user interface (GUI) module, which gradually displays the cards using a typewriter effect, and also provides it to the voice broadcast module, which broadcasts the new response. Therefore, when the backtracking dialogue content is incomplete, generating new response content based on the dialogue question can provide complete answers to dialogue questions that the user previously did not receive, improving the user experience.

[0257] In some feasible embodiments, during voice broadcasting, the type of service for the current question and answer is indicated by setting facial expressions for the current broadcast content. For example, for entertainment-related questions and answers, cheerful and relaxed facial expressions, such as those for listening to music, can be set; for encyclopedic questions and answers, serious and formal facial expressions, such as those for carrying teaching aids or wearing glasses, can be set. Similarly, facial expressions are used to indicate the service type of the current dialogue during dialogue rewinding; specific facial expressions can be set according to user needs.

[0258] Dialogue backtracking is equivalent to a normal dialogue interaction, so it also needs to synchronize context information to the natural language understanding module for semantic arbitration. During the dialogue backtracking process, the business notification interaction module caches and applies the context.

[0259] Each time a non-dialogue rewind intent is received from a large module, the interaction module is notified to cache the response content for that round; each time a dialogue rewind intent is received, the interaction module is notified to add the cached content from the previous round to the context list.

[0260] A fourth aspect of this application provides an electronic device, comprising:

[0261] The memory is configured to store instructions; and

[0262] The processor is configured to retrieve the instructions from the memory and to implement the dialogue method when executing the instructions.

[0263] A fifth aspect of this application provides a vehicle including a control system that applies the aforementioned dialogue method.

[0264] A sixth aspect of this application provides a machine-readable storage medium storing instructions that cause a machine to perform the aforementioned dialogue method.

[0265] A seventh aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the dialogue method.

[0266] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0267] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0268] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0269] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0270] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0271] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0272] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0273] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0274] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A dialogue method, characterized in that, The dialogue method includes: When a dialogue request is received, and the dialogue request includes at least one of several semantic keywords that are used to continue, extend, supplement, or expand upon the previous dialogue, the dialogue request triggers dialogue backtracking. When a dialogue rewind is triggered, a dialogue rewind strategy is determined based on the dialogue request and the stored key dialogue information. The key dialogue information only includes information from the previous round of dialogue, including: dialogue content, dialogue content completeness, and dialogue progress information. The dialogue content includes the first response content. If no dialogue content is available for backtracking, a new dialogue will begin; If there is dialogue content for backtracking, and the first response content has been generated and output, then the dialogue question and the first response content of the previous round are obtained from the key dialogue information, and supplementary response content is generated based on the dialogue question and the first response content. If there is dialogue content for backtracking, and the first response content has been generated but not yet output, then the first response content and dialogue progress information are obtained from the key dialogue information. If there is dialogue content for backtracking, and the first response content has not been fully generated, then the dialogue question from the previous round is obtained from the key dialogue information, and a new response content is generated based on the dialogue question. Based on the dialogue backtracking strategy, the target response content corresponding to the dialogue request is output, including: The system can either broadcast the response to a new conversation via voice and display it in the same way as a new conversation, or broadcast the supplementary content of the response via voice and display it in the same way as a new conversation, or continue to broadcast the first response via voice based on the conversation progress information and display it in the same way as a new conversation, or broadcast the new response via voice and display it in the same way as a new conversation.

2. The dialogue method according to claim 1, characterized in that, The step of determining a dialogue backtracking strategy based on the dialogue request and stored key dialogue information includes: Determine if there is key information in the dialogue to determine if there is dialogue content that can be used for backtracking; Based on the completeness of the dialogue content, determine whether the first response content has been fully generated; Based on the dialogue progress information, determine whether the first response content has been completely output; The dialogue backtracking strategy is determined based on whether there is dialogue content for backtracking, whether the first response content has been generated, and whether the first response content has been output.

3. The dialogue method according to claim 1, characterized in that, The key information of the dialogue is stored in the following manner: When generating response content at the start of the current round of dialogue, clear the key dialogue information of the previous round of dialogue from the cache, and store the key dialogue information of the current round of dialogue in the cache.

4. The dialogue method according to claim 1, characterized in that, The dialogue backtracking strategy includes: a first dialogue backtracking strategy, a second dialogue backtracking strategy, a third dialogue backtracking strategy, and a fourth dialogue backtracking strategy; After determining the dialogue backtracking strategy, when the dialogue backtracking strategy is the first dialogue backtracking strategy, the key dialogue information of the new dialogue is stored in the cache. When the dialogue backtracking strategy is the second dialogue backtracking strategy, when generating supplementary response content, the key dialogue information of the previous round of dialogue in the cache is cleared, and the dialogue request and the generated supplementary response content are stored in the cache. When the dialogue backtracking strategy is the third dialogue backtracking strategy, after obtaining the first reply content from the dialogue key information, the dialogue key information of the previous round of dialogue in the cache is cleared, the dialogue request is stored in the cache, and the first reply content is stored in the cache as the target reply content. When the dialogue backtracking strategy is the fourth dialogue backtracking strategy, when generating new response content, the key dialogue information of the previous round of dialogue in the cache is cleared, and the dialogue request and the generated new response content are stored in the cache.

5. The dialogue method according to claim 1, characterized in that, The step of outputting the target response content corresponding to the dialogue request according to the dialogue backtracking strategy includes: The dialogue backtracking strategy is executed to obtain the target response content; Based on the target response content, output the target response content according to the output strategy; The output strategy includes a voice broadcast strategy and a response content display strategy.

6. The dialogue method according to claim 5, characterized in that, The voice broadcast strategy includes guidance information output and target response content output, and the guidance information output and the target response content output are linked by listening to the end signal of the guidance information output. The guidance information includes a new dialogue start prompt voice when the first dialogue backtracking strategy is executed, and an original dialogue continuation prompt voice when the third or fourth dialogue backtracking strategy is executed.

7. The dialogue method according to claim 1, characterized in that, After receiving the dialogue request but before determining the dialogue backtracking strategy, the method further includes: Identify the intent behind the dialogue request; Determine whether the stated intent is the intent of the target vertical domain; When the intent is the intent of the target vertical domain, determine whether the dialogue request triggers dialogue backtracking; When a dialogue backtracking is triggered, execute the steps to determine the dialogue backtracking strategy.

8. The dialogue method according to claim 1, characterized in that, The dialogue method also includes: Upon receiving a signal that the conversation has been interrupted, the interruption signal is synchronized to the voice avatar, which then provides a follow-up prompt.

9. A dialogue device, characterized in that, The dialogue device includes: A data receiving unit is used to receive a dialogue request. When the dialogue request includes at least one of several semantic keywords that are used to continue, extend, supplement, or expand upon the previous dialogue, the dialogue request triggers dialogue backtracking. The strategy determination unit is used to determine a dialogue backtracking strategy based on the dialogue request and the stored key dialogue information; the key dialogue information only includes information from the previous round of dialogue, and the key dialogue information includes: dialogue content, dialogue content completeness and dialogue progress information, and the dialogue content includes the first response content; If no dialogue content is available for backtracking, a new dialogue will begin; If there is dialogue content for backtracking, and the first response content has been generated and output, then the dialogue question and the first response content of the previous round are obtained from the key dialogue information, and supplementary response content is generated based on the dialogue question and the first response content. If there is dialogue content for backtracking, and the first response content has been generated but not yet output, then the first response content and dialogue progress information are obtained from the key dialogue information. If there is dialogue content for backtracking, and the first response content has not been fully generated, then the dialogue question from the previous round is obtained from the key dialogue information, and a new response content is generated based on the dialogue question. The execution unit is configured to output the target response content corresponding to the dialogue request according to the dialogue backtracking strategy, including: The system can either broadcast the response to a new conversation via voice and display it in the same way as a new conversation, or broadcast the supplementary content of the response via voice and display it in the same way as a new conversation, or continue to broadcast the first response via voice based on the conversation progress information and display it in the same way as a new conversation, or broadcast the new response via voice and display it in the same way as a new conversation.

10. A dialogue system, characterized in that, The system includes: a natural language understanding module, a chat application module, a large model module, a cache, and an output module; The natural language understanding module is used to understand the dialogue request and send the understood user intent to the chat application module. When the dialogue request includes at least one of several semantic keywords that are used to continue, extend, supplement, or expand on the previous dialogue, the dialogue request triggers dialogue backtracking. The chat application module is used to query the stored key dialogue information from the cache according to the user's intent, and determine the dialogue backtracking strategy according to the key dialogue information; the key dialogue information only contains information from the previous round of dialogue, and the key dialogue information includes: dialogue content, dialogue content completeness and dialogue progress information, and the dialogue content includes the first reply content; The chat application module is also used to call the large model module and / or the natural language understanding module according to the dialogue backtracking strategy to determine the target response content corresponding to the dialogue request; If there is dialogue content for backtracking, and the first response content has been generated and output, then the chat application module requests the natural language understanding module to continue generating semantics. The natural language understanding module identifies the semantics of the previous round of dialogue questions and sends it to the big model module. The big model module generates supplementary response content based on the semantics of the previous round of dialogue questions and sends it to the chat application module. If there is dialogue content for backtracking, and the first reply content has been generated but not yet output, then the chat application module obtains the first reply content as the target reply content; If there is dialogue content for backtracking, and the first response content has not been fully generated, the chat application module requests the previous round of dialogue questions from the natural language understanding module. The natural language understanding module identifies the semantics of the previous round of dialogue questions and sends it to the big model module. The big model module generates new response content based on the semantics of the previous round of dialogue questions and sends it to the chat application module. The output module is used to receive the target reply content sent by the chat application module and output the target reply content; The output module includes a voice broadcast module and a graphical interface module; If the dialogue backtracking strategy determined by the chat application module is the second dialogue backtracking strategy, then the chat application module sends the supplementary reply content to the graphical interface module for display, and at the same time sends the supplementary reply content to the voice broadcasting module for broadcasting. If the dialogue backtracking strategy determined by the chat application module is the third dialogue backtracking strategy, then the chat application module will send the target reply content to the graphical interface module for display, and at the same time send it to the voice broadcasting module to broadcast the target reply content based on the dialogue progress information. If the dialogue backtracking strategy determined by the chat application module is the fourth dialogue backtracking strategy, then the chat application module will send the new reply content to the graphical interface module for display, and at the same time send the new reply content to the voice broadcasting module for broadcasting.

11. An electronic device, characterized in that, include: The memory is configured to store instructions; as well as The processor is configured to retrieve the instructions from the memory and, when executing the instructions, to implement the dialogue method of any one of claims 1 to 8.

12. A vehicle, comprising a control system, characterized in that, The control system applies the dialogue method as described in any one of claims 1 to 8.

13. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores instructions for causing the machine to perform the dialogue method of any one of claims 1 to 8.

14. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the dialogue method as described in any one of claims 1 to 8.

Images