Voice interaction method and device of vehicle, equipment and medium

Abstract

The invention discloses a voice interaction method and device for a vehicle, equipment and a medium, and the method comprises the steps: converting received user voice data into a user query text, extracting query key information from the user query text, and transmitting the query key information to a content provider; obtaining information source content which is fed back by the content provider and corresponds to the query key information, and extracting reply key information from the information source content; and obtaining user features and scene features of the current vehicle, generating a target broadcast text according to the user features, the scene features and the reply key information, converting the target broadcast text into broadcast voice data, and broadcasting the broadcast voice data through a vehicle-mounted loudspeaker. According to the method and the device, the reply key information is extracted from the information source content fed back by the content provider, so that the problems of redundant voice reply information and relatively long time consumption can be effectively avoided, and personalized voice interaction can be realized by combining the user characteristics and the scene characteristics when the target broadcast text is generated.

Description

Technical Field

[0001] The present invention relates to the field of vehicle technology, and in particular to a vehicle voice interaction method, device, terminal equipment and computer-readable storage medium. Background Art

[0002] With the continuous development of vehicle technology, the application of voice interaction technology in automobiles has become more and more common. Users can easily obtain various information through voice interaction technology while driving, such as weather information, oil price information, road congestion, etc. Existing voice interaction methods usually send user query text to content providers and directly broadcast based on the content returned by the content providers, making voice reply information redundant, time-consuming, and lacking in personalization. Summary of the invention

[0003] The present invention provides a vehicle voice interaction method, device, equipment and medium. By extracting key information for replying to source content fed back by a content provider, the problems of redundant voice reply information and long time consumption can be effectively avoided. When generating a target broadcast text, personalized voice interaction can be achieved by combining user characteristics and scene characteristics.

[0004] In order to solve the above technical problems, a first aspect of an embodiment of the present invention provides a vehicle voice interaction method, comprising the following steps:

[0005] Converting the received user voice data into user query text, extracting key query information from the user query text and sending it to the content provider;

[0006] Acquire the source content corresponding to the query key information fed back by the content provider, and extract the reply key information from the source content;

[0007] The user characteristics and scene characteristics of the current vehicle are obtained, and a target broadcast text is generated according to the user characteristics, the scene characteristics and the key reply information, and the target broadcast text is converted into broadcast voice data and broadcasted through the vehicle-mounted speakers.

[0008] As a preferred solution, generating the target broadcast text according to the user characteristics, the scene characteristics and the reply key information specifically includes the following steps:

[0009] Based on a preset broadcast text template database, a target broadcast text template is obtained from the broadcast text template database according to the user characteristics, the scene characteristics and the reply key information;

[0010] The target broadcast text is generated according to the reply key information and the target broadcast text template.

[0011] As a preferred solution, generating the target broadcast text according to the user characteristics, the scene characteristics and the reply key information specifically includes the following steps:

[0012] The user features, the scene features and the key reply information are input into a pre-trained large language model, and the target broadcast text is generated using the pre-trained large language model.

[0013] As a preferred embodiment, the method further comprises the following steps:

[0014] Based on the extracted key reply information, determining whether the key reply information meets a preset scenario response condition;

[0015] When the reply key information satisfies the scenario response condition, based on a preset scenario response database, a target response scenario, a plurality of vehicle execution modules corresponding to the target response scenario, and execution actions corresponding to each of the vehicle execution modules are determined from the scenario response database according to the user characteristics, the scenario characteristics, and the reply key information;

[0016] According to the execution actions corresponding to the vehicle execution modules, control instructions corresponding to the vehicle execution modules are generated, and the control instructions are sent to the vehicle execution modules.

[0017] As a preferred solution, the step of judging whether the reply key information meets the preset scenario response condition specifically includes the following steps:

[0018] When the reply key information contains preset trigger information corresponding to any response scenario, or the numerical information in the reply key information reaches the preset trigger numerical range corresponding to any response scenario, it is determined that the reply key information meets the scenario response condition; wherein the scenario response database includes several response scenarios.

[0019] As a preferred solution, extracting key query information from the user query text and sending it to the content provider specifically includes the following steps:

[0020] Performing intent recognition and named entity recognition on the user query text respectively to determine the intent type of the user query text and a number of entities in the user query text;

[0021] The intent type and the plurality of entities are sent to the content provider.

[0022] As a preferred solution, extracting the key reply information from the source content specifically includes the following steps:

[0023] Intent recognition and named entity recognition are performed on the source content respectively to determine the intent type of the source content and several entities in the source content.

[0024] As a preferred solution, the user characteristics at least include age, gender, preference information and vehicle use habit information;

[0025] The scene features include at least vehicle driving status, geographic location and time.

[0026] A second aspect of an embodiment of the present invention provides a voice interaction device for a vehicle, comprising:

[0027] A primary semantic understanding module, used to convert received user voice data into user query text, extract key query information from the user query text and send it to the content provider;

[0028] A secondary semantic understanding module, used to obtain the information source content corresponding to the query key information fed back by the content provider, and extract the reply key information from the information source content;

[0029] The voice broadcast module is used to obtain the user characteristics and scene characteristics of the current vehicle, and generate a target broadcast text according to the user characteristics, the scene characteristics and the reply key information, convert the target broadcast text into broadcast voice data and broadcast it through the vehicle-mounted speaker.

[0030] A third aspect of an embodiment of the present invention provides a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the voice interaction method for the vehicle as described in any one of the first aspects is implemented.

[0031] A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the vehicle voice interaction method as described in any one of the first aspects.

[0032] Compared with the prior art, the beneficial effect of the embodiments of the present invention is that by extracting key reply information from the source content fed back by the content provider, the problems of redundant voice reply information and long time consumption can be effectively avoided. When generating the target broadcast text, personalized voice interaction can be achieved by combining user characteristics and scene characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a flow chart of a voice interaction method for a vehicle in an embodiment of the present invention;

[0034] Figure 2 is a diagram of the voice interaction architecture of a vehicle in an embodiment of the present invention;

[0035] Figure 3 It is a flowchart of a preferred embodiment of the voice interaction scenario provided by the present invention;

[0036] Figure 4 It is a flowchart of another preferred embodiment of the voice interaction scenario provided by the present invention;

[0037] Figure 5 It is a flowchart of another preferred embodiment of the voice interaction scenario provided by the present invention;

[0038] Figure 6 It is a flowchart of another preferred embodiment of the voice interaction scenario provided by the present invention;

[0039] Figure 7 It is a schematic diagram of the structure of a voice interaction device for a vehicle in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] See also Figure 1 and Figure 2 According to a first aspect of an embodiment of the present invention, a vehicle voice interaction method is provided, comprising the following steps S1 to S3:

[0042] Step S1, converting the received user voice data into user query text, extracting key query information from the user query text and sending it to the content provider;

[0043] Step S2, obtaining the source content corresponding to the query key information fed back by the content provider, and extracting the reply key information from the source content;

[0044] Step S3, obtaining the user characteristics and scene characteristics of the current vehicle, and generating a target broadcast text according to the user characteristics, the scene characteristics and the reply key information, converting the target broadcast text into broadcast voice data and broadcasting it through the vehicle-mounted speaker.

[0045] Specifically, this embodiment first receives user voice data through an audio receiving device such as a microphone, and performs voice recognition on the received user voice data to obtain user query text. Further, semantic recognition is performed on the user query text, and query key information is extracted therefrom and sent to the content provider, so that the content provider provides corresponding source content based on the query key information. For example, if the query key information is weather, the source content fed back by the content provider includes but is not limited to location, weather conditions, wind direction, temperature and air quality; if the query key information is oil price, the source content fed back by the content provider includes but is not limited to the local retail price of No. 92 gasoline, No. 95 gasoline, No. 98 gasoline and diesel on that day.

[0046] It is worth noting that different content providers may provide different information sources. Some content providers may provide information sources with too much redundant information, which may result in a long voice response time, and most of the response content may not be of concern to the user. For example, a user asks: "How is the oil price today?" The voice response directly based on the information source content is: "Today's retail price of No. 92 gasoline: 8.03 yuan / liter, No. 95 gasoline: 8.59 yuan / liter, No. 98 gasoline: 9.61 yuan / liter. The actual oil price is subject to the local gas station." However, the user is only concerned about the price of No. 92 gasoline, and the expected response should be: "The No. 92 gasoline you care about is 8.03 yuan / liter today." The user asks: "How is the weather today?" The voice response directly based on the information source content is: "Shanghai, today is sunny, northeast wind, 33℃~39℃, the current temperature is 38 degrees, and the air quality is excellent." However, the user is only concerned about the highest temperature, and the expected response should be: "Today is very hot, the highest temperature is 39℃, please be careful to prevent heat stroke." Therefore, in order to avoid redundant voice reply information and time-consuming, the present embodiment extracts key reply information from the source content and obtains the user characteristics and scene characteristics of the current vehicle, so as to generate personalized target broadcast text based on the current user characteristics and scene characteristics, and convert the target broadcast text into broadcast voice data through the speech synthesis module and broadcast it through the vehicle-mounted speakers, so as to realize voice broadcast that is different for each person.

[0047] As one of the optional embodiments, in order to improve the recognition accuracy of the user query text and thus enable accurate voice interaction, this embodiment performs front-end processing on the user voice data after receiving the user voice data.

[0048] As a preferred solution, the user characteristics at least include age, gender, preference information and vehicle use habit information; the scene characteristics at least include vehicle driving status, geographic location and time. It is worth noting that the user characteristics in this embodiment include but are not limited to directly obtaining through user account information, image recognition through camera, voiceprint recognition, etc. Generally speaking, the vehicle account service stores user-related personal information, including age, gender, preference information and vehicle use habit information, etc., and the relevant user characteristics can be directly obtained by accessing the vehicle account service. When the vehicle account is switched, the user characteristics are also switched accordingly. In addition, the current passenger image in the car is captured by the camera, and the target detection task is performed using a neural network, so that the current user characteristics, including age, gender, etc., can be obtained. In addition, the current user characteristics, including age, gender, etc., can also be determined by voiceprint recognition; the scene characteristics include but are not limited to obtaining through the vehicle system service or the vehicle scene engine, or directly reading the current vehicle driving status, geographic location and time and other scene characteristics.

[0049] A vehicle voice interaction method provided by an embodiment of the present invention can effectively avoid the problems of redundant voice reply information and long time consumption by extracting key reply information from the source content fed back by the content provider. When generating the target broadcast text, personalized voice interaction can be achieved by combining user characteristics and scene characteristics.

[0050] As a preferred solution, generating the target broadcast text according to the user characteristics, the scene characteristics and the reply key information specifically includes the following steps:

[0051] Based on a preset broadcast text template database, a target broadcast text template is obtained from the broadcast text template database according to the user characteristics, the scene characteristics and the reply key information;

[0052] The target broadcast text is generated according to the reply key information and the target broadcast text template.

[0053] Specifically, the broadcast text generated by this embodiment is implemented based on hard-coded logic, which is a solidified code. All possible broadcast text templates under different user characteristics and scene characteristics are set according to actual voice interaction needs to form a broadcast text template database. When the current user characteristics, scene characteristics and reply key information are obtained, the corresponding target broadcast text template can be obtained from the broadcast text template database, and then the reply key information is combined with the target broadcast text template to generate the target broadcast text, thereby realizing differentiated text output based on different user characteristics and scene characteristics.

[0054] As a preferred solution, generating the target broadcast text according to the user characteristics, the scene characteristics and the reply key information specifically includes the following steps:

[0055] The user features, the scene features and the key reply information are input into a pre-trained large language model, and the target broadcast text is generated using the pre-trained large language model.

[0056] It is worth noting that the broadcast text generated by the present embodiment can also be implemented based on a large language model trained with preliminary data. It can be understood that the present embodiment pre-trains the large language model using a broadcast text training set marked with user features, scene features and reply key information. The large language model adjusts model parameters by learning the user features, scene features and reply key information of each training broadcast text. In actual application, the user features, scene features and reply key information are input into the pre-trained large language model, and the pre-trained large language model can be used to generate personalized broadcast text.

[0057] As an optional embodiment, the pre-trained large language model is specifically a GPT large model.

[0058] As a preferred embodiment, the method further comprises the following steps:

[0059] Based on the extracted key reply information, determining whether the key reply information meets a preset scenario response condition;

[0060] When the reply key information satisfies the scenario response condition, based on a preset scenario response database, a target response scenario, a plurality of vehicle execution modules corresponding to the target response scenario, and execution actions corresponding to each of the vehicle execution modules are determined from the scenario response database according to the user characteristics, the scenario characteristics, and the reply key information;

[0061] According to the execution actions corresponding to the vehicle execution modules, control instructions corresponding to the vehicle execution modules are generated, and the control instructions are sent to the vehicle execution modules.

[0062] It is worth noting that in order to enhance the intelligent experience of human-computer interaction, this embodiment can also call the vehicle's hardware and software devices to perform related actions based on the reply key information to achieve three-dimensional interactive feedback. Specifically, when it is determined that the reply key information meets the preset scene response conditions, based on the preset scene response database, according to user characteristics, scene characteristics and reply key information, the target response scene, the vehicle execution modules corresponding to the target response scene, and the execution actions corresponding to each vehicle execution module are determined from the scene response database.

[0063] It can be understood that the scene response database contains several preset response scenes, each response scene is set with corresponding user characteristics, scene characteristics and reply key information as trigger conditions. In addition, each response scene is also provided with a corresponding vehicle execution module and the execution actions corresponding to each vehicle execution module. Exemplarily, the vehicle execution module includes but is not limited to a robotic arm module, a fragrance module, a lighting and sound effect module, a seat module and a central control screen module. The execution actions corresponding to the vehicle execution module include but are not limited to the swinging of the robotic arm, the swinging of the robotic arm, the opening of the fragrance channel by the fragrance module, the turning on of the atmosphere light, the playing of rain sound effects, the playing of thunder sound effects, the swinging of the virtual image displayed on the central control screen, etc.

[0064] Furthermore, a corresponding control instruction is generated according to the determined execution action corresponding to each vehicle execution module, and each vehicle execution model is controlled so that each vehicle execution model executes the corresponding execution action based on the control instruction.

[0065] As a preferred solution, the step of judging whether the reply key information meets the preset scenario response condition specifically includes the following steps:

[0066] When the reply key information contains preset trigger information corresponding to any response scenario, or the numerical information in the reply key information reaches the preset trigger numerical range corresponding to any response scenario, it is determined that the reply key information meets the scenario response condition; wherein the scenario response database includes several response scenarios.

[0067] Specifically, when there is preset trigger information corresponding to any response scenario in the reply key information, for example, the preset trigger information corresponding to a rainy day response scenario is weather description information related to rain, and the current reply key information contains "moderate to heavy rain", then the reply key information is determined to meet the scenario response conditions of the rainy day response scenario; or, when the numerical information in the reply key information reaches the preset trigger numerical range corresponding to any response scenario, for example, the preset trigger numerical range corresponding to a stock rise response scenario is that the stock rise is greater than 1%, and the current reply key information contains "rise" and "1.25%", then the reply key information is determined to meet the scenario response conditions of the stock rise response scenario.

[0068] For example, Figures 3 to 6 As shown in the figure, four personalized voice interaction scenarios are provided in this embodiment. Figure 3In the video, a user asks: "Did Moutai go up today?" The source content is: "The current share price of Kweichow Moutai is 1875.00, down -15.28, a drop of -0.81%." The key reply information extracted from it is "share price 1875.00", "down" and "0.81%." The target broadcast text generated is: "Alas, it has fallen again. The current share price is 1875.00, a drop of -0.81%." Based on the user's pre-set response scenario of controlling the central control screen to perform a downward swing action when the stock falls, in order to reflect a sigh, the corresponding control instruction is generated to control the central control screen to perform a downward swing action. Figure 4 In the example, a user asks: "How is the market today?" The source content is: "The Shanghai Composite Index is now 3360.62, up 58.21, an increase of 1.25%." The key information extracted from the reply is "index 3360.62", "up" and "1.25%." The target report text is generated: "Wow, the market rose 1.25 points today, and the current index is 3360.62." Based on the user's pre-settings to open the fragrance channel when the stock rises, control the central control screen to swing up and down and left and right, and control the color of the atmosphere light to change in a jumpy manner, the corresponding control instructions are generated to control the fragrance module to open the fragrance channel, control the central control screen to swing up and down and left and right, and control the atmosphere light module to turn on the atmosphere light for a preset time and the color of the atmosphere light changes in a jumpy manner. Figure 5 In the example, a user asks: "What will the weather be like tomorrow?" The source content is: "Beijing will have moderate to heavy rain tomorrow, northeast wind, 13℃~21℃, air quality index 99, and good air quality." The key response information extracted from it is "moderate to heavy rain," "13℃," and "21℃." Based on the time in the scene feature: tomorrow is Wednesday, the user preference information in the user features includes whether it will rain and the temperature. Therefore, the target broadcast text generated is: "It will have moderate to heavy rain tomorrow and it will be a bit cold. Remember to bring an umbrella to work." Based on the key response information, the corresponding target response scene is filtered out from the scene response database. The vehicle execution module includes a central control screen module and a lighting and sound effect module. Therefore, the corresponding control instructions are generated to control the virtual image in the central control screen to open an umbrella and to control the lighting and sound effect module to play thunder sound effects. Figure 6 In the example, a user asks: "How much is the oil price now?" The source content is: "Today's retail price of No. 92 gasoline in Anhui Province is 8.03 yuan / liter, the retail price of No. 95 gasoline is 8.59 yuan / liter, and the retail price of No. 98 gasoline is 9.61 yuan / liter. The actual oil price is subject to the local gas station." The key information for the reply is extracted: "Retail price of No. 92 gasoline: 8.03 yuan / liter", "Retail price of No. 95 gasoline: 8.59 yuan / liter" and "Retail price of No. 98 gasoline: 9.61 yuan / liter". Based on the car usage habit information in the user's characteristics, including the use of No. 92 gasoline, the target broadcast text is generated: "The No. 92 gasoline you usually use is 8.03 yuan / liter today."

[0069] As a preferred solution, extracting key query information from the user query text and sending it to the content provider specifically includes the following steps:

[0070] Performing intent recognition and named entity recognition on the user query text respectively to determine the intent type of the user query text and a number of entities in the user query text;

[0071] The intent type and the plurality of entities are sent to the content provider.

[0072] Specifically, natural language understanding mainly performs semantic recognition on user query text. In voice interaction, it is necessary to understand the user's intention in a sentence and extract the keywords that the user cares about, that is, slots. Based on the NLU (Natural Language Understanding) model, this embodiment performs intent recognition and named entity recognition on user query text. It can be understood that this embodiment pre-sets several intent types and several slots corresponding to each intent type. Through intent recognition, the user query text can be classified into the corresponding intent type, and then based on the several slots corresponding to the intent type, the user query text is subjected to named entity recognition, and several entities in the user query text are extracted. For example, suppose a user asks: "Has Moutai increased today?" Through intent recognition, the current query is distinguished as the stock field, the intent type is stock price query, and the corresponding slots include time and stock. Through named entity recognition, the entities in the user query text are extracted: time: "today", stock: "Moutai"; suppose a user asks: "Will it rain today?" Through intent recognition, it is determined that the current intent type is weather query, and the corresponding slots include time and weather type. Through named entity recognition, the entities in the user query text are extracted: time: "today", weather type: "raining". Further, the intent type and several extracted entities are sent to the content provider so that the content provider can accurately provide the corresponding source content.

[0073] As a preferred solution, extracting the key reply information from the source content specifically includes the following steps:

[0074] Intent recognition and named entity recognition are performed on the source content respectively to determine the intent type of the source content and several entities in the source content.

[0075] Specifically, since the same entity may represent different entity types in different intent types, for example, assuming that a user asks what the price of Moutai is on a shopping platform, the intent type at this time is product price query, and the entity type corresponding to "Moutai" is the product name. Suppose that the user asks whether Moutai has increased today, the intent type at this time is stock price query, and the entity type corresponding to "Moutai" is the stock name. Therefore, in order to ensure the accuracy of the voice reply, this embodiment needs to perform intent recognition and named entity recognition on the source content to determine its intent type and several entities therein.

[0076] It is worth noting that when the broadcast text is implemented based on hard-coded logic, each broadcast text template in the preset broadcast text template database is set with a corresponding intent type and / or entity trigger condition. For example, when the user preference information in the user characteristics includes the temperature, different temperature ranges can be set to correspond to different broadcast text templates. For example, if the temperature is lower than 15°C, the broadcast text template contains a reply text such as "It's a bit cold"; if the temperature is between 15°C and 28°C, the broadcast text template contains a reply text such as "The temperature is moderate"; if the temperature is higher than 28°C, the broadcast text template contains a reply text such as "It's a bit hot". Therefore, according to user characteristics, scene characteristics, intent types and several entities, the corresponding target broadcast text template can be accurately matched from the broadcast text template database, and then the target broadcast text is generated based on several entities and the target broadcast text template.

[0077] When the broadcast text is implemented based on a large language model trained with preliminary data, as mentioned above, the same entity may represent different entity types in different intent types, and thus the corresponding broadcast texts may also be different. Therefore, in the process of generating the target broadcast text using the pre-trained large language model, using the determined intent type as the model input can ensure that the pre-trained large language model accurately generates the target broadcast text, thereby ensuring the accuracy of the voice response.

[0078] As an optional embodiment, Figure 2 The two natural language processing services and natural language generation in the shown architecture can be uniformly implemented by a pre-trained large language model.

[0079] See also Figure 7 A second aspect of an embodiment of the present invention provides a voice interaction device for a vehicle, comprising:

[0080] A primary semantic understanding module 701, used to convert received user voice data into user query text, extract key query information from the user query text and send it to the content provider;

[0081] A secondary semantic understanding module 702 is used to obtain the source content corresponding to the query key information fed back by the content provider, and extract the reply key information from the source content;

[0082] The voice broadcast module 703 is used to obtain the user characteristics and scene characteristics of the current vehicle, and generate a target broadcast text according to the user characteristics, the scene characteristics and the reply key information, convert the target broadcast text into broadcast voice data and broadcast it through the vehicle-mounted speaker.

[0083] As a preferred solution, the voice broadcast module 703 is used to generate a target broadcast text according to the user characteristics, the scene characteristics and the reply key information, specifically including:

[0084] Based on a preset broadcast text template database, a target broadcast text template is obtained from the broadcast text template database according to the user characteristics, the scene characteristics and the reply key information;

[0085] The target broadcast text is generated according to the reply key information and the target broadcast text template.

[0086] As a preferred solution, the voice broadcast module 703 is used to generate a target broadcast text according to the user characteristics, the scene characteristics and the reply key information, specifically including:

[0087] The user features, the scene features and the key reply information are input into a pre-trained large language model, and the target broadcast text is generated using the pre-trained large language model.

[0088] As a preferred solution, the device further includes a scene response module, which is used to:

[0089] Based on the extracted key reply information, determining whether the key reply information meets a preset scenario response condition;

[0090] When the reply key information satisfies the scenario response condition, based on a preset scenario response database, a target response scenario, a plurality of vehicle execution modules corresponding to the target response scenario, and execution actions corresponding to each of the vehicle execution modules are determined from the scenario response database according to the user characteristics, the scenario characteristics, and the reply key information;

[0091] According to the execution actions corresponding to the vehicle execution modules, control instructions corresponding to the vehicle execution modules are generated, and the control instructions are sent to the vehicle execution modules.

[0092] As a preferred solution, the scenario response module is used to determine whether the reply key information meets the preset scenario response conditions, specifically including:

[0093] When the reply key information contains preset trigger information corresponding to any response scenario, or the numerical information in the reply key information reaches the preset trigger numerical range corresponding to any response scenario, it is determined that the reply key information meets the scenario response condition; wherein the scenario response database includes several response scenarios.

[0094] As a preferred solution, the primary semantic understanding module 701 is used to extract key query information from the user query text and send it to the content provider, specifically including:

[0095] Performing intent recognition and named entity recognition on the user query text respectively to determine the intent type of the user query text and a number of entities in the user query text;

[0096] The intent type and the plurality of entities are sent to the content provider.

[0097] As a preferred solution, the secondary semantic understanding module 702 is used to extract key reply information from the source content, specifically including:

[0098] Intent recognition and named entity recognition are performed on the source content respectively to determine the intent type of the source content and several entities in the source content.

[0099] As a preferred solution, the user characteristics at least include age, gender, preference information and vehicle use habit information;

[0100] The scene features include at least vehicle driving status, geographic location and time.

[0101] It should be noted that a vehicle voice interaction device provided in an embodiment of the present invention can implement all processes of the vehicle voice interaction method described in any of the above embodiments. The functions of each module in the device and the technical effects achieved are respectively the same as the functions and technical effects achieved by the vehicle voice interaction method described in the above embodiments, and will not be repeated here.

[0102] A third aspect of an embodiment of the present invention provides a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the voice interaction method for a vehicle as described in any embodiment of the first aspect is implemented.

[0103] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory. The terminal device may also include an input / output device, a network access device, a bus, etc.

[0104] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0105] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the terminal device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0106] A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the vehicle voice interaction method as described in any embodiment of the first aspect.

[0107] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus a necessary hardware platform, and of course can also be implemented entirely by hardware. Based on such an understanding, all or part of the contribution of the technical solution of the present invention to the background technology can be embodied in the form of a software product, and the computer software product can be stored in a storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present invention or certain parts of the embodiments.

[0108] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A vehicle voice interaction method, characterized in that: The steps include: Converting the received user voice data into user query text, extracting key query information from the user query text and sending it to the content provider; Acquire the source content corresponding to the query key information fed back by the content provider, and extract the reply key information from the source content; The user characteristics and scene characteristics of the current vehicle are obtained, and a target broadcast text is generated according to the user characteristics, the scene characteristics and the key reply information, and the target broadcast text is converted into broadcast voice data and broadcasted through the vehicle-mounted speakers.

2. The vehicle voice interaction method according to claim 1, characterized in that: The generating of the target broadcast text according to the user characteristics, the scene characteristics and the key reply information specifically comprises the following steps: Based on a preset broadcast text template database, a target broadcast text template is obtained from the broadcast text template database according to the user characteristics, the scene characteristics and the reply key information; The target broadcast text is generated according to the reply key information and the target broadcast text template.

3. The vehicle voice interaction method according to claim 1, characterized in that: The generating of the target broadcast text according to the user characteristics, the scene characteristics and the key reply information specifically comprises the following steps: The user features, the scene features and the key reply information are input into a pre-trained large language model, and the target broadcast text is generated using the pre-trained large language model.

4. The vehicle voice interaction method according to claim 1, characterized in that: The method further comprises the steps of: Based on the extracted key reply information, determining whether the key reply information meets a preset scenario response condition; When the reply key information satisfies the scenario response condition, based on a preset scenario response database, a target response scenario, a plurality of vehicle execution modules corresponding to the target response scenario, and execution actions corresponding to each of the vehicle execution modules are determined from the scenario response database according to the user characteristics, the scenario characteristics, and the reply key information; According to the execution actions corresponding to the vehicle execution modules, control instructions corresponding to the vehicle execution modules are generated, and the control instructions are sent to the vehicle execution modules.

5. The vehicle voice interaction method as claimed in claim 4, characterized in that: The step of judging whether the reply key information meets the preset scenario response condition specifically includes the following steps: When the reply key information contains preset trigger information corresponding to any response scenario, or the numerical information in the reply key information reaches the preset trigger numerical range corresponding to any response scenario, it is determined that the reply key information meets the scenario response condition; wherein the scenario response database includes several response scenarios.

6. The vehicle voice interaction method according to claim 1, characterized in that: The step of extracting key query information from the user query text and sending the information to the content provider specifically includes the following steps: Performing intent recognition and named entity recognition on the user query text respectively to determine the intent type of the user query text and a number of entities in the user query text; The intent type and the plurality of entities are sent to the content provider.

7. The vehicle voice interaction method according to claim 1, characterized in that: The step of extracting the key reply information from the source content specifically includes the following steps: Intent recognition and named entity recognition are performed on the source content respectively to determine the intent type of the source content and several entities in the source content.

8. The vehicle voice interaction method according to any one of claims 1 to 7, characterized in that: The user characteristics include at least age, gender, preference information and vehicle use habit information; The scene features include at least vehicle driving status, geographic location and time.

9. A voice interaction device for a vehicle, characterized in that: include: A primary semantic understanding module, used to convert received user voice data into user query text, extract key query information from the user query text and send it to the content provider; A secondary semantic understanding module, used to obtain the information source content corresponding to the query key information fed back by the content provider, and extract the reply key information from the information source content; The voice broadcast module is used to obtain the user characteristics and scene characteristics of the current vehicle, and generate a target broadcast text according to the user characteristics, the scene characteristics and the reply key information, and convert the target broadcast text into broadcast voice data and broadcast it through the vehicle-mounted speaker.

10. A terminal device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the voice interaction method for a vehicle as described in any one of claims 1 to 8 when executing the computer program.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the voice interaction method for a vehicle as described in any one of claims 1 to 8.

Images