Voice interaction method, server and readable storage medium

Abstract

The invention discloses a voice interaction method, a server and a nonvolatile computer readable storage medium containing a computer program. The method comprises the steps of receiving a first voice request forwarded by a vehicle; according to the first voice request, a target intention, target intention parameters and a target intention parameter filling result are obtained, a target intention set is constructed in advance, the target intention set comprises multiple target intentions, and each target intention is configured with an associated target intention parameter. And determining target feedback content according to the target intention, the target intention parameter and the target intention parameter filling result. And issuing the target feedback content to the vehicle to guide completion of voice interaction. Therefore, through the pre-constructed target intention set, the system can quickly understand and respond to the voice request of the user, so that the dependence on a predefined dialogue script can be reduced, more flexible and personalized interaction experience can be provided, and efficient task-based dialogue can be realized.

Description

Technical Field

[0001] The present invention relates to the field of voice interaction technology, and in particular to a voice interaction method, a server and a non-volatile computer-readable storage medium containing a computer program. Background Art

[0002] In the related art, the dialogue system usually uses a multi-round dialogue script to guide users to complete the voice interaction task. However, the dialogue system requires a large amount of labeled data, the constructed dialogue system has poor portability, and the cost is high. Summary of the invention

[0003] The present application provides a voice interaction method, a server, and a non-volatile computer-readable storage medium containing a computer program.

[0004] The present application provides a voice interaction method, the method comprising:

[0005] receiving a first voice request forwarded by a vehicle;

[0006] According to the first voice request, a target intent, a target intent parameter and a target intent parameter filling result are obtained, wherein a target intent set is pre-built, the target intent set includes a plurality of the target intents, and each of the target intents is configured with the associated target intent parameter;

[0007] Determining target feedback content according to the target intent, the target intent parameter, and the target intent parameter filling result;

[0008] The target feedback content is sent to the vehicle to guide the completion of the voice interaction.

[0009] In this way, the server receives the first voice request forwarded by the vehicle. Then, the server obtains the target intent, target intent parameters and target intent parameter filling results according to the first voice request, wherein the target intent set is pre-built, the target intent set includes multiple target intents, and each target intent is configured with associated target intent parameters. Then, the server determines the target feedback content according to the target intent, target intent parameters and target intent parameter filling results. Finally, the server sends the target feedback content to the vehicle to guide the completion of the voice interaction. In this way, through the pre-built target intent set, the system can quickly understand and respond to the user's voice request, thereby reducing the dependence on predefined dialogue scripts, providing a more flexible and personalized interactive experience, and realizing efficient task-based dialogue. In addition, by using the pre-built target intent set and target intent parameters, the dependence on a large amount of annotated data is reduced. The dialogue system can quickly adapt to new tasks and fields without or with only a small amount of annotated data, thereby reducing the cost of data collection and annotation.

[0010] In some implementations, obtaining a target intent, a target intent parameter, and a target intent parameter filling result according to the first voice request includes:

[0011] Performing natural language recognition processing on the first voice request to obtain a natural language recognition result;

[0012] Based on a preset large language model, the target intent and the target intent parameters are determined according to the natural language recognition result and a pre-built task database;

[0013] Natural language understanding is performed on the natural language recognition result, and target intent parameter filling is performed according to the natural language understanding result to obtain the target intent parameter filling result.

[0014] In this way, the server performs natural language recognition processing on the first voice request to obtain a natural language recognition result. Then, based on the preset large language model, the server determines the target intent and target intent parameters according to the natural language recognition result and the pre-built task database. Finally, the server performs natural language understanding on the natural language recognition result, and performs target intent parameter filling according to the natural language understanding result to obtain the target intent parameter filling result. In this way, based on the preset large language model and the pre-built task database, the server can more accurately determine the user's target intent and target intent parameters, and determine how to respond to guide the user to complete the voice interaction. Moreover, by performing the target intent parameter filling, the server can quickly prepare all the necessary information and parameters, thereby reducing the user's waiting time and improving the efficiency of task completion. In addition, by using a pre-built task database and a large language model, the dialogue system can more easily adapt to new tasks and fields without a large amount of data annotation and model training, which facilitates the rapid deployment and expansion of the dialogue system.

[0015] In some implementations, the target intent parameter filling result includes filled and unfilled, and determining the target feedback content according to the target intent, the target intent parameter and the target intent parameter filling result includes:

[0016] When the target intent parameters are all filled, the application interface corresponding to the target intent is selected to perform processing according to the target intent parameter filling result and the target intent to obtain the target feedback content.

[0017] In this way, when all the target intent parameters are filled, the server selects the application interface corresponding to the target intent to execute the processing according to the target intent parameter filling result and the target intent, and obtains the target feedback content. In this way, by selecting and executing the corresponding application interface when all the target intent parameters are filled, the server can improve the accuracy of task execution, enhance user experience, and improve the response speed of the system.

[0018] In some implementations, determining target feedback content according to the target intent, the target intent parameter, and the target intent parameter filling result includes:

[0019] In the case where there is at least one unfilled target intent parameter, the target feedback content is determined based on the unfilled target intent parameter.

[0020] In this way, when there is at least one unfilled parameter in the target intent, the server determines the target feedback content based on the unfilled target intent parameter. In this way, by identifying the unfilled parameters and directly asking the user for relevant information, the server can obtain the required information more quickly, thereby reducing the need for multiple rounds of interaction and improving the overall interaction efficiency. In addition, through targeted inquiries, it helps users clarify the information they need to provide, reduces the cognitive burden of users during the interaction process, makes it easier for users to understand and respond to system requests, and improves user experience.

[0021] In some implementations, sending the target feedback content to the vehicle to guide the completion of the voice interaction includes:

[0022] In the case where there are at least two unfilled parameters in the target intention parameter, the target feedback contents are sent to the vehicle one by one to guide the user to issue a second voice request for each target feedback content one by one.

[0023] In this way, when there are at least two unfilled parameters in the target intent parameter, the server sends the target feedback content to the vehicle one by one to guide the user to issue a second voice request for each target feedback content one by one. In this way, by sending the target feedback content one by one, the server can guide the user to supplement the missing information step by step, thereby helping the user to understand and respond to the system's requests more easily, thereby improving the overall interaction efficiency and accuracy. In addition, the server can dynamically generate feedback content based on the unfilled parameters and adjust the interaction process based on the user's response, so that the system can adapt to different user requests and scenarios, improving the flexibility and adaptability of the system.

[0024] In certain embodiments, the method further comprises:

[0025] Receive the second voice request forwarded by the vehicle to complete the voice interaction.

[0026] In this way, the server receives the second voice request forwarded by the vehicle to complete the voice interaction. In this way, by guiding the user to supplement the missing target intention parameters, the system can reduce erroneous operations caused by misunderstanding or errors and improve the accuracy and reliability of the system.

[0027] In certain embodiments, the method further comprises:

[0028] Storing the first voice request and the second voice request in a preset database;

[0029] According to the first voice request and the second voice request, the target intent parameter filling result is calibrated.

[0030] In this way, the server stores the first voice request and the second voice request in a preset database. Then, the server calibrates the target intent parameter filling result according to the first voice request and the second voice request. In this way, by matching and calibrating the target intent parameter filling result in real time through the first voice request and the second voice request, the system can quickly adapt to the user's input and context changes, thereby providing a more accurate response. Moreover, through real-time matching and calibration, the server can output the rule template and intervene as soon as a problem occurs, thereby reducing misunderstandings and errors in the interaction process and improving the reliability of the system.

[0031] In certain embodiments, the method further comprises:

[0032] When the first voice request or the second voice request includes a preset detection word, the voice interaction is stopped.

[0033] In this way, when the first voice request or the second voice request includes the preset detection word, the server stops the voice interaction. In this way, by recognizing the preset detection word, the voice interaction is terminated or paused, so that the user has a sense of control over the interaction process, and the user experience is improved. In addition, by recognizing the preset detection word, the dialogue can be quickly reset to avoid further errors.

[0034] An embodiment of the present application provides a server, which includes a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the voice interaction method as described above is implemented.

[0035] An embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the voice interaction method as described above is implemented.

[0036] Additional aspects and advantages of the embodiments of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0038] Figure 1 It is one of the flowcharts of the voice interaction method of certain embodiments of the present application;

[0039] Figure 2 is a schematic diagram of a voice request processing flow in certain implementation modes of the present application;

[0040] Figure 3 This is the second flow chart of the voice interaction method of certain implementation modes of the present application;

[0041] Figure 4 This is the third flow chart of the voice interaction method of certain implementation modes of the present application;

[0042] Figure 5 This is a fourth flowchart of a voice interaction method according to certain embodiments of the present application;

[0043] Figure 6 This is a fifth flow chart of a voice interaction method according to certain embodiments of the present application;

[0044] Figure 7 This is the sixth flow chart of the voice interaction method of certain implementation modes of the present application;

[0045] Figure 8 It is a schematic diagram of the missing target intention parameter collection process of certain embodiments of the present application;

[0046] Fig. 9 This is the seventh flow chart of the voice interaction method of certain implementation modes of the present application;

[0047] Fig.10 It is a schematic diagram of the processing flow of the rule system of certain embodiments of the present application;

[0048] Fig.11 This is the eighth flow chart of the voice interaction method of certain embodiments of the present application. DETAILED DESCRIPTION

[0049] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions from beginning to end. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the embodiments of the present application, and cannot be understood as limiting the embodiments of the present application.

[0050] In the field of artificial intelligence, a dialogue system is a system designed to interact with users through natural language. They can be applied in a variety of scenarios, such as customer service, smart home control, virtual assistants, etc. In these applications, task-based dialogue systems are particularly important because they are designed to help users complete specific tasks, such as booking a flight, setting an alarm, or querying information.

[0051] Traditional task-based dialogue systems usually rely on multi-turn dialogue scripts. These scripts are pre-defined dialogue flows that guide users through a series of interactive steps to complete a specific task. For example, in the scenario of booking a flight, the dialogue system may ask the user where they want to go, when they want to leave, what class they need, etc., and then complete the task based on the user's answers.

[0052] However, traditional task-based dialogue systems that rely on multi-round dialogue scripts have some significant disadvantages. First, building and maintaining these multi-round dialogue scripts requires a lot of time and human resources. Each dialogue task may require a unique script, and when the tasks or requirements change, these scripts also need to be updated accordingly. Second, this traditional task-based dialogue system has poor transferability. That is, a dialogue system built for a specific task is difficult to reuse for other tasks because the dialogue flow for each task may be different. Finally, this approach requires a large amount of annotated data to train the dialogue system so that it can understand the user's intentions and responses. Obtaining this data can be time-consuming and expensive.

[0053] Based on the above questions, please refer to Figure 1 , the present application embodiment provides a voice interaction method, the method comprising:

[0054] 01: receiving the first voice request forwarded by the vehicle;

[0055] 02: According to the first voice request, obtain the target intent, target intent parameters and target intent parameter filling result;

[0056] 03: Determine the target feedback content according to the target intent, target intent parameters and target intent parameter filling results;

[0057] 04: Send the target feedback content to the vehicle to guide the completion of voice interaction.

[0058] The embodiment of the present application also provides a server, including a memory and a processor. The voice interaction method of the embodiment of the present application can be implemented by the server of the embodiment of the present application. Specifically, a computer program is stored in the memory, and the processor is used to receive a first voice request forwarded by the vehicle. And according to the first voice request, a target intent, a target intent parameter, and a target intent parameter filling result are obtained. The processor is also used to determine the target feedback content according to the target intent, the target intent parameter, and the target intent parameter filling result. And the target feedback content is sent to the vehicle to guide the completion of the voice interaction.

[0059] The embodiment of the present application also provides a voice interaction device. The voice interaction method of the embodiment of the present application can be implemented by the voice interaction device of the embodiment of the present application. Specifically, the voice interaction device includes a receiving module, a language processing module, a determination module and a sending module. The receiving module is used to receive a first voice request forwarded by the vehicle. The language processing module is used to obtain the target intent, target intent parameters and target intent parameter filling results according to the first voice request. The determination module is used to determine the target feedback content according to the target intent, target intent parameters and target intent parameter filling results. The sending module is used to send the target feedback content to the vehicle to guide the completion of the voice interaction.

[0060] Specifically, the first voice request refers to the first voice request input by the user when interacting with the server. The first voice request is the starting point for the server to understand and respond to the user's intention, which triggers a series of interactive processes to help the user complete a specific task. For example, in a vehicle application scenario, the user may say "navigate to the nearest gas station", and this voice input is the first voice request.

[0061] The target intent refers to the specific operation or task that the system recognizes the user wants to perform based on the user's voice request. For example, if the user's voice request is "navigate to the nearest gas station", the recognized target intent is "navigation".

[0062] "Goal intent parameters" refer to the parameters associated with a specific goal intent. These parameters include specific information required to complete the goal intent, which are key components of task execution. For example, if the user's goal intent is "navigation", then the goal intent parameters may include "destination" (such as location A) and "route preference" (such as shortest distance or fewest traffic lights, etc.).

[0063] The target intent parameter filling result refers to the result after the system identifies and fills in the specific parameters required for the target intent based on the user's voice request. The target intent parameter filling result includes all the detailed information required to complete the target intent, which are the key inputs for task execution. For example, if the user's target intent is "navigation", and the relevant target intent parameters include "destination" and "route preference", then the target intent parameter filling result may include "destination: location A" and "route preference: shortest distance". These target intent parameter filling results will be used to guide the system to perform corresponding operations, such as navigating to a specific location and selecting a route preferred by the user.

[0064] The target intent set refers to a pre-built list of intents, which are a collection of all possible user requests that the system can understand and process. Each target intent represents a specific operation or task that a user may want to perform, such as "make a call", "navigate to a destination", "play music", etc.

[0065] Target feedback content refers to feedback information determined by the system based on the user's voice request, target intent, target intent parameters and their filling results. Target feedback content is the system's response to the user's request, used to guide the vehicle or other equipment to complete the user's request, or when the system lacks the target intent parameters necessary to complete the voice interaction, it is sent to guide the user to supplement the corresponding target intent parameter information. For example, if the user's target intent is "navigation", and the relevant target intent parameter filling results include "Destination: Location A" and "Route Preference: Shortest Distance", then the target feedback content may be "The shortest distance route to Location A has been searched, and you are heading to Location A". If the user's target intent is "navigation", and the relevant target intent parameter filling result is "Destination: Location A", then the target feedback content may be "The route to Location A has been searched, please select your route preference".

[0066] See also Figure 2 , Figure 2 This is a diagram of the voice request processing process. First, the R&D personnel build a task database, taking the example of "I want to initiate navigation":

[0067] (1) Configure the application interface (i.e., intent) for initiating navigation

[0068] (2) Configure the information required to call the application interface, such as destination and route preference (i.e., intent parameters).

[0069] Second, push the configuration information to the configuration platform. Third, interact with the user based on the large language model to obtain the information required to execute the application interface. Fourth, the rule system is responsible for all process monitoring and final result verification to prevent the large model illusion from causing task failure. Fifth, finally integrate the data and call the task application interface to complete the voice interaction.

[0070] After the server receives the first voice request forwarded by the vehicle, it accurately identifies the user's first voice request, quickly determines the target intent and target intent parameters from the pre-built target intent set, and then confirms the target intent parameter filling result. In this way, the server can efficiently understand the user's actual needs and provide accurate feedback content, so that the vehicle can quickly respond to the user's instructions and complete the corresponding tasks, such as navigation, playing music, or adjusting the temperature in the car.

[0071] Subsequently, the server determines the information that needs to be fed back to the vehicle, namely, the target feedback content, based on the target intent, target intent parameters, and target intent parameter filling results.

[0072] Finally, the server sends the target feedback content to the vehicle to guide the completion of the voice interaction.

[0073] In summary, in the voice interaction method and server provided by the embodiment of the present application, the server receives a first voice request forwarded by the vehicle. Then, the server obtains the target intent, target intent parameters and target intent parameter filling results according to the first voice request, wherein the target intent set is pre-built, the target intent set includes multiple target intents, and each target intent is configured with associated target intent parameters. Then, the server determines the target feedback content according to the target intent, target intent parameters and target intent parameter filling results. Finally, the server sends the target feedback content to the vehicle to guide the completion of the voice interaction. In this way, through the pre-built target intent set, the system can quickly understand and respond to the user's voice request, thereby reducing the dependence on predefined dialogue scripts, providing a more flexible and personalized interactive experience, and realizing efficient task-based dialogue. In addition, by using the pre-built target intent set and target intent parameters, the dependence on a large amount of annotated data is reduced. The dialogue system can quickly adapt to new tasks and fields without or with only a small amount of annotated data, thereby reducing the cost of data collection and annotation.

[0074] See also Figure 3 In some implementations, step 02 (obtaining a target intent, a target intent parameter, and a target intent parameter filling result according to the first voice request) includes:

[0075] 021: Perform natural language recognition processing on the first voice request to obtain a natural language recognition result;

[0076] 022: Based on the preset large language model, the target intent and target intent parameters are determined according to the natural language recognition results and the pre-built task database;

[0077] 023: Perform natural language understanding on the natural language recognition result, and perform target intent parameter filling according to the natural language understanding result to obtain the target intent parameter filling result.

[0078] In some embodiments, the language processing module is further used to perform natural language recognition processing on the first voice request to obtain a natural language recognition result. And based on a preset large language model, determine the target intent and target intent parameters according to the natural language recognition result and a pre-built task database. And perform natural language understanding on the natural language recognition result, and perform target intent parameter filling according to the natural language understanding result to obtain a target intent parameter filling result.

[0079] In some embodiments, the processor is further configured to perform natural language recognition processing on the first voice request to obtain a natural language recognition result. Based on a preset large language model, the target intent and target intent parameters are determined according to the natural language recognition result and a pre-built task database. The natural language recognition result is subjected to natural language understanding, and target intent parameter filling is performed according to the natural language understanding result to obtain a target intent parameter filling result.

[0080] Specifically, the preset large language model refers to a pre-trained natural language processing model, usually a deep learning model such as Transformer or BERT, which can be used to understand and interpret the user's voice request, convert the voice into text, and extract useful information and intent from it. It should be noted that the preset large language model will directly output the target intent and will not output additional information. If the target intent cannot be determined, the reply will be undetermined.

[0081] The pre-built task database refers to a database that stores all possible tasks that the system can process and their related information. Each task in this database has a target intent, and each target intent has a set of target intent parameters associated with it, which are required to successfully execute the target intent. In some embodiments, an application programming interface is directly used as the target intent, that is, the application programming interface that needs to be called to implement the task. The task database may include the following information:

[0082] Pre_info=[{"API":"Navigation","AF":["Destination","Route Preference"]},

[0083] {"API":"Book a hotel","AF":["Check-in time","Check-out time","City"]},

[0084] {"API":"Call Phone","AF":["Contact Person"]}]

[0085] Among them, API represents the application programming interface that needs to be called to implement the task (ie, the target intent), and AF represents the parameters required to successfully execute the API.

[0086] It should be noted that when the server determines the target intent and target intent parameters, the large language model can directly output the target intent and target intent parameters at the same time, or a discriminant model can first identify the target intent and then obtain the corresponding target intent parameters from the task database based on the target intent.

[0087] Target intent parameter filling refers to the process of identifying and filling specific parameters related to the target intent based on the user's voice request. For example, if the user's voice request is "navigate to the nearest gas station", then the target intent parameter filling is to fill "nearest gas station" into the target intent parameter "destination".

[0088] Automatic Speech Recognition (ASR) processing refers to the technology of converting human speech signals into machine-understandable text or commands.

[0089] Natural Language Understanding (NLU) refers to the technologies and processes that enable computers to understand, interpret, and generate human language.

[0090] The server first performs automatic speech recognition processing on the vehicle's voice request and converts the voice signal into a natural language recognition result in text form. Then, the server uses a preset large language model, combined with the natural language recognition results and a pre-built task database, to determine the user's target intent and the target intent parameters associated with it. Finally, the server further understands the natural language recognition results in depth. Based on the results of natural language understanding, the server performs the filling of target intent parameters, that is, according to the user's request content, the specific parameter values ​​are filled into the corresponding intent parameters, thereby obtaining a complete intent parameter filling result.

[0091] For example, if the user's voice request is "I want to initiate navigation to the nearest gas station", the server may perform the following steps:

[0092] First, ASR: Converting Voice Requests to Text I want to initiate navigation to the nearest gas station.

[0093] Second, target intent and target intent parameter identification: The server uses a large language model and task database to identify that the user’s target intent is “navigation” and determines that the required target intent parameters include “destination” and “route preference”.

[0094] Third, NLU: The server understands that the user’s destination is the nearest gas station.”

[0095] Fourth, target intent parameter filling: the server fills the destination parameter with “nearest gas station”.

[0096] The corresponding prompt of the target intention discrimination model is as follows:

[0097] If you are a discriminant model, you need to help me determine which of the following intentions the user input belongs to?

[0098] Intent list: "_""+",.join(APIs)+""_".

[0099] The "_""+",.join(APIs)+""_" is used to convert a list (such as ["navigation", "call"]) into a string (such as "navigation, call"). The list refers to the target intent obtained based on the user's voice request. In this way, you can easily use this string for output or display, for example, to list all possible intents in the discriminant model.

[0100] The target intent parameter identifies the corresponding system prompt setting as follows:

[0101] Suppose you are a voice conversation assistant and your existing business information is as follows:

[0102] Business indicator API: "_""+api_name+""_"

[0103] Conditions required to execute the API ("_""+",.join(AF)+""_")

[0104] Among them, the role of the business indicator API: "_""+api_name+""_" is to prompt the big model system what the target business interface of the current session is, that is, the target intent. api_name is a variable that represents the specific business interface name, such as "navigation" or "call". The role of the conditions required to execute the API ("_""+",.join(AF)+""_") is to prompt the big model system with all the conditions (target intent parameters) required to execute the current business interface (api_name). AF is a list containing all the necessary parameter names, such as ["destination", "route preference"]. ",".join(AF) is a Python expression used to connect the elements in the AF list into a comma-delimited string. In this way, the big model system knows what specific information needs to be collected from the user. By including these conditions in the prompt, the big model system knows what questions to ask the user in multiple rounds of conversations, and how to construct the JSON output format so that the corresponding target intent parameter information can be updated after the user enters information.

[0105] In this way, based on the preset large language model and the pre-built task database, the server can more accurately determine the user's target intent and target intent parameters, and determine how to respond to guide the user to complete the voice interaction. And, by performing target intent parameter filling, the server can quickly prepare all necessary information and parameters, thereby reducing the user's waiting time and improving the efficiency of task completion. In addition, by using a pre-built task database and large language model, the dialogue system can more easily adapt to new tasks and fields without a large amount of data annotation and model training, which facilitates the rapid deployment and expansion of the dialogue system.

[0106] See also Figure 4 In some implementations, the target intent parameter filling result includes filled and unfilled, and step 03 (determining the target feedback content according to the target intent, the target intent parameter and the target intent parameter filling result) includes:

[0107] 031: When all target intent parameters are filled, select the application interface corresponding to the target intent to execute processing according to the target intent parameter filling result and the target intent to obtain the target feedback content.

[0108] In some embodiments, the determination module is also used to select an application interface corresponding to the target intent to execute processing based on the target intent parameter filling results and the target intent when all target intent parameters are filled in, so as to obtain target feedback content.

[0109] In some embodiments, the processor is also used to, when all target intent parameters are filled, select an application interface corresponding to the target intent to execute processing based on the target intent parameter filling results and the target intent to obtain target feedback content.

[0110] Specifically, after the server fills in the target intent parameters according to the first voice request, the model derives from the user voice request that the target intent parameters for executing the intent have been filled in. Subsequently, the server selects the application interface corresponding to the target intent based on the filled target intent parameters and the target intent. For example, if the target intent is "navigation", the server will select the navigation application interface.

[0111] Next, the server calls the selected application interface and passes the filled target intent parameters. For example, the navigation application interface plans the navigation route according to the passed target intent parameters, that is, plans the navigation route according to "destination: location A" and "route selection: least traffic lights".

[0112] Finally, after the application interface performs processing, it generates target feedback content. For example, the navigation application interface generates target feedback content "A route to the nearest gas station has been planned for you, please follow the instructions."

[0113] In this way, by selecting and executing the corresponding application interface when all target intent parameters are filled in, the server can improve the accuracy of task execution, enhance user experience, and increase the response speed of the system.

[0114] See also Figure 5 In some embodiments, step 03 (determining target feedback content according to target intent, target intent parameters, and target intent parameter filling results) includes:

[0115] 032: When there is at least one unfilled target intent parameter, the target feedback content is determined based on the unfilled target intent parameter.

[0116] In some embodiments, the determination module is further configured to determine the target feedback content based on the unfilled target intent parameters when at least one of the target intent parameters is unfilled.

[0117] In some embodiments, the processor is further configured to determine target feedback content based on the unfilled target intent parameters when at least one of the target intent parameters is unfilled.

[0118] Specifically, after the server fills in the target intent parameter according to the first voice request, the model concludes according to the user's voice request that there is still one target intent parameter that has not been filled in for executing the intent. Subsequently, the server analyzes the user's voice request and identifies which target intent parameters have not been filled in. For example, if the user's request is "Navigate to Location A" but no specific place name is provided, the target intent parameter "Route Preference" is not filled in.

[0119] Next, the server generates targeted feedback content based on the unfilled target intent parameters to guide the user to provide the missing information. For example, the server may generate targeted feedback content as "Please tell me what your route preference is?" Or, the user requests to play music in the in-vehicle system, but does not specify a specific song or artist. The server recognizes that the song or artist parameter is unfilled, and then determines the targeted feedback content as "Please tell me which song or artist you want to play?".

[0120] In this way, by identifying unfilled parameters and directly asking users for relevant information, the server can obtain the required information more quickly, thereby reducing the need for multiple rounds of interaction and improving overall interaction efficiency. In addition, through targeted inquiries, it helps users clarify the information they need to provide, reducing the cognitive burden of users during the interaction process, making it easier for users to understand and respond to system requests, and improving user experience.

[0121] See also Figure 6 In some implementations, step 04 (sending the target feedback content to the vehicle to guide the completion of the voice interaction) includes:

[0122] 041: When there are at least two unfilled target intent parameters, the target feedback contents are sent to the vehicle one by one to guide the user to issue a second voice request for each target feedback content one by one.

[0123] In some embodiments, the sending module is also used to send the target feedback content to the vehicle one by one when there are at least two unfilled target intent parameters, so as to guide the user to issue a second voice request for each target feedback content one by one.

[0124] In some embodiments, the processor is also used to send the target feedback content to the vehicle one by one when there are at least two unfilled target intent parameters, so as to guide the user to issue a second voice request for each target feedback content one by one.

[0125] Specifically, after the server fills in the target intent parameters according to the first voice request, the model infers based on the user's voice request that there are at least two unfilled target intent parameters for executing the intent. It will generate target feedback content for each unfilled parameter in turn and send them to the vehicle one by one to guide the user to provide the missing information one by one.

[0126] For example, the user's voice request is "Help me navigate", but no specific place name and route preference are provided. The server recognizes that the destination and route preference parameters are not filled in, and then determines two target feedback contents: "Please tell me the destination you want to navigate to?" and "Please tell me which route you want to choose?".

[0127] The server sends the first target feedback content to the vehicle. After the user responds, the server processes the voice request in response to the user to confirm whether the voice request in response can correctly obtain the missing target intent parameter "destination". If the missing target intent parameter "destination" can be obtained correctly, the server will send the second target feedback content and continue to guide the user to provide the missing target intent parameter information "route preference". Finally, the server fills in the target intent parameter according to the information provided by the user and performs the corresponding operation. If the missing target intent parameter "destination" cannot be obtained correctly, the server will continue to guide the user to provide the first missing target intent parameter "destination".

[0128] In this way, by sending targeted feedback content one by one, the server can guide users to fill in the missing information step by step, thereby helping users to understand and respond to system requests more easily, thereby improving the overall interaction efficiency and accuracy. In addition, the server can dynamically generate feedback content based on unfilled parameters and adjust the interaction process based on the user's response, so that the system can adapt to different user requests and scenarios, improving the flexibility and adaptability of the system.

[0129] See also Figure 7 In some embodiments, the method further comprises:

[0130] 05: Receive the second voice request forwarded by the vehicle to complete the voice interaction.

[0131] In some embodiments, the receiving module is further used to receive a second voice request forwarded by the vehicle to complete the voice interaction.

[0132] In some embodiments, the processor is further configured to receive a second voice request forwarded by the vehicle to complete the voice interaction.

[0133] Specifically, the second voice request refers to the target intent parameter information supplemented by the user based on the target feedback content sent by the server. This is the additional information that the server guides the user to provide when some target intent parameters are missing in the first voice request.

[0134] See also Figure 8 , Figure 8 Schematic diagram of the process of collecting missing target intent parameters. As shown in the figure, after the server receives the first voice request sent by the user, the large language model understands the first voice request based on the pre-built task database, and obtains the target intent, target intent parameters, and target intent parameter filling results. In this process, the rule system in the server will continue to calibrate the understanding process of the large language model. If the rule system in the server finds that some target intent parameters are not filled, the large language model will generate target feedback content and send it to the vehicle, requesting the user to provide the missing information. The user provides the missing information through voice based on the target feedback content sent by the server.

[0135] For example, a user sends a voice request in the vehicle system, "Help me navigate and choose a route with fewer traffic lights," without providing a specific location name. The server generates the target feedback content "Please tell me where you want to navigate to?" and sends it to the vehicle. The user responds, "I want to go to the nearest supermarket." This response is the second voice request, which adds the target intent parameter "destination."

[0136] In this way, by guiding users to supplement missing target intent parameters, the system can reduce erroneous operations due to misunderstandings or errors and improve the accuracy and reliability of the system.

[0137] See also Fig. 9 In some embodiments, the method further comprises:

[0138] 06: Storing the first voice request and the second voice request in a preset database;

[0139] 07: According to the first voice request and the second voice request, calibrate the target intent parameter filling result.

[0140] In some embodiments, the voice interaction device further includes a calibration module, which is used to store the first voice request and the second voice request in a preset database, and to perform calibration processing on the target intent parameter filling result according to the first voice request and the second voice request.

[0141] In some embodiments, the processor is further configured to store the first voice request and the second voice request in a preset database, and to perform calibration processing on the target intention parameter filling result according to the first voice request and the second voice request.

[0142] Specifically, the preset database refers to a database used by the server to store and manage voice requests and related data, including historical voice requests of users, detailed information of the requests, processing results, etc.

[0143] Calibration refers to the process in which the rule system in the server verifies and adjusts the target intent parameter filling results according to the content of the first voice request and the second voice request, which can ensure the accuracy and consistency of parameter filling, so that the system can respond to the user's request more accurately. Verification refers to the server checking whether the filled parameters are consistent with the user's actual request. Adjustment refers to the server making corresponding adjustments if errors or inconsistencies are found in the parameter filling. See Fig.10 , Fig.10 This is a processing flow diagram of the rule system. The rule system is responsible for all process monitoring and final result verification to prevent the large model from hallucinating and causing task failure. Among them, after the large language model receives the forwarded voice request, it performs preset detection word detection processing on the voice request to confirm whether the voice request includes the preset detection word. If the voice request includes the preset detection word, the current round of dialogue will be exited. If the voice request does not include the preset detection word, the rule system will calibrate the dialogue data generated by the large language model in real time according to the pre-built task database. When it is confirmed that some target intent parameters are not filled in, a guide word is generated based on the unfilled target intent parameter, and the target feedback content is generated based on the pre-examination prompt word template and the guide word.

[0144] The server stores the user's first voice request and the second voice request in a preset database to record the user's complete request history for subsequent analysis and optimization. Then, the server verifies and adjusts the target intent parameter filling results based on the content of the first voice request and the second voice request to ensure the accuracy and consistency of the target intent parameter filling, so that the system can respond to the user's request more accurately.

[0145] For example, a user makes a voice request in the vehicle system, "Help me navigate and choose a route with fewer traffic lights", without providing a specific location name. The server generates the target feedback content "Please tell me where you want to navigate to?" and sends it to the vehicle. The server generates the target feedback content "Please tell me where you want to navigate to?" and sends it to the vehicle. The user responds, "I want to go to the nearest supermarket." This response is the second voice request, which adds the target intent parameter "destination".

[0146] The server receives this second voice request, extracts the destination parameter information "the nearest supermarket", and then continues to perform the navigation task. In this human-computer interaction process, the server will check the conversation records stored in the preset database to confirm whether the filled parameters are consistent with the user's actual request. If errors or inconsistencies are found, corresponding adjustments will be made. For example, if the supermarket mentioned by the user does not exist in the system, the server may ask the user again for the exact location name or provide similar locations for the user to choose.

[0147] In this way, by matching and calibrating the target intent parameter filling results in real time through the first voice request and the second voice request, the system can quickly adapt to the user's input and context changes, thereby providing a more accurate response. In addition, through real-time matching and calibration, the server can output the rule template and intervene as soon as a problem occurs, thereby reducing misunderstandings and errors in the interaction process and improving the reliability of the system.

[0148] See also Fig.11 In some embodiments, the method further comprises:

[0149] 08: When the first voice request and the second voice request include a preset detection word, stop the voice interaction.

[0150] In some embodiments, the voice interaction device further includes an exit module, which is used to stop the voice interaction when the first voice request and the second voice request include a preset detection word.

[0151] In some embodiments, the processor is further configured to stop the voice interaction when the first voice request and the second voice request include a preset detection word.

[0152] Specifically, the preset detection words refer to some keywords or phrases pre-set in the server. When these words or phrases appear in the user's first voice request or second voice request, the server will recognize them and stop the current voice interaction accordingly, allowing the user to easily exit when no further interaction is required. For example, "Exit", "No thanks", "Cancel", etc.

[0153] Please refer to Fig.10 ,When the server detects that the user’s voice request contains a preset detection word, it will immediately stop the current voice interaction process and will no longer ask or perform subsequent operations.

[0154] For example, the user's voice request in the car system is "request to play music", but then changes his mind and says the preset detection word "no thanks". After the server recognizes the word "no thanks", it will immediately stop the current voice interaction and no longer ask the user which song or artist he wants to play.

[0155] In this way, by recognizing the preset detection words, the voice interaction is terminated or paused, so that the user has a sense of control over the interaction process, which improves the user experience. In addition, by recognizing the preset detection words, the conversation can be quickly reset to avoid further errors.

[0156] The present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method of the present application is executed.

[0157] It is understood that a computer program includes computer program code. The computer program code may be in source code form, object code form, executable file or some intermediate form. Computer readable storage media may include: any entity or device capable of carrying computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), and software distribution medium.

[0158] In the description of this specification, the descriptions with reference to the terms "specifically", "further", "particularly", "understandably", etc. are intended to mean that the specific features, structures, materials or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms are not intended to refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0159] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0160] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A voice interaction method, characterized in that: include: receiving a first voice request forwarded by a vehicle; According to the first voice request, a target intent, a target intent parameter and a target intent parameter filling result are obtained, wherein a target intent set is pre-built, the target intent set includes a plurality of the target intents, and each of the target intents is configured with the associated target intent parameter; Determining target feedback content according to the target intent, the target intent parameter, and the target intent parameter filling result; The target feedback content is sent to the vehicle to guide the completion of the voice interaction.

2. The voice interaction method according to claim 1, characterized in that: The obtaining, according to the first voice request, a target intent, a target intent parameter, and a target intent parameter filling result, includes: Performing natural language recognition processing on the first voice request to obtain a natural language recognition result; Based on a preset large language model, the target intent and the target intent parameters are determined according to the natural language recognition result and a pre-built task database; Natural language understanding is performed on the natural language recognition result, and target intent parameter filling is performed according to the natural language understanding result to obtain the target intent parameter filling result.

3. The voice interaction method according to claim 1, characterized in that: The target intent parameter filling result includes filled and unfilled, and determining the target feedback content according to the target intent, the target intent parameter and the target intent parameter filling result includes: When the target intent parameters are all filled, the application interface corresponding to the target intent is selected to perform processing according to the target intent parameter filling result and the target intent to obtain the target feedback content.

4. The voice interaction method according to claim 3, characterized in that: The determining target feedback content according to the target intent, the target intent parameter and the target intent parameter filling result includes: In the case where there is at least one unfilled target intent parameter, the target feedback content is determined based on the unfilled target intent parameter.

5. The voice interaction method according to claim 4, characterized in that: The sending of the target feedback content to the vehicle to guide the completion of the voice interaction includes: In the case where there are at least two unfilled parameters in the target intention parameter, the target feedback contents are sent to the vehicle one by one to guide the user to issue a second voice request for each target feedback content one by one.

6. The voice interaction method according to claim 5, characterized in that: The method further comprises: Receive the second voice request forwarded by the vehicle to complete the voice interaction.

7. The voice interaction method according to claim 5, characterized in that: The method further comprises: Storing the first voice request and the second voice request in a preset database; According to the first voice request and the second voice request, the target intent parameter filling result is calibrated.

8. The voice interaction method according to claim 5, characterized in that: The method further comprises: When the first voice request or the second voice request includes a preset detection word, the voice interaction is stopped.

9. A server, characterized in that: The server includes a processor and a memory, and a computer program is stored in the memory. When the computer program is executed by the processor, the voice interaction method according to any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the voice interaction method as described in any one of claims 1 to 8 is implemented.

Images