Voice interaction method and apparatus, device, medium, and product

Abstract

A voice interaction method and apparatus, a device, a medium, and a product, relating to the technical field of artificial intelligence, in particular to the technical fields of automatic driving, large language models, intelligent cabins and the like. The method comprises: acquiring voice information corresponding to a target user collected by a vehicle-mounted terminal (11), and acquiring text information corresponding to the voice information (201); on the basis of the text information, determining a text expression type corresponding to the voice information (202); determining understanding result information of the voice information by using a target language model corresponding to the text expression type (203); and outputting the understanding result information by means of the vehicle-mounted terminal (11) (204).

Description

Voice interaction method, device, equipment, medium and product

[0001] This disclosure claims priority to the Chinese patent application filed with the China Patent Office on September 28, 2023, with application number 202311280914.3 and application name “Voice Interaction Methods, Devices, Equipment, Media and Products,” the entire contents of which are incorporated by reference into this disclosure. Technical Field

[0002] The present disclosure relates to technical fields such as autonomous driving, large language models, and smart cockpits in the field of artificial intelligence, and in particular to a voice interaction method, device, equipment, medium, and product. Background Art

[0003] A car computer is an in-vehicle information device installed in a car that enables communication between drivers, between vehicles, and between vehicles and backend servers. Currently, a car computer can collect voice messages from the driver, interpret the corresponding voice messages, and then output the interpretation. This interpretation can, for example, display navigation information or initiate phone calls.

[0004] However, the current car computer can only handle simple voice interaction scenarios. For some complex voice interaction scenarios, the accuracy of the understanding results is poor, resulting in voice interaction failure.

[0005] Summary of the Invention

[0006] The present disclosure provides a voice interaction method, apparatus, device, medium and product for human-computer interaction scenarios.

[0007] According to a first aspect of the present disclosure, a voice interaction method is provided, comprising:

[0008] Acquire the voice information corresponding to the target user collected by the vehicle terminal, and obtain the text information corresponding to the voice information;

[0009] Determining, based on the text information, a text expression type corresponding to the voice information;

[0010] Determining understanding result information of the speech information by using a target language model corresponding to the text expression type;

[0011] The understanding result information is outputted through the vehicle-mounted terminal.

[0012] According to a second aspect of the present disclosure, a voice interaction device is provided, comprising:

[0013] An information acquisition unit, configured to acquire voice information corresponding to a target user collected by the vehicle-mounted terminal, and to acquire text information corresponding to the voice information;

[0014] a type determination unit, configured to determine a text expression type corresponding to the voice information based on the text information;

[0015] a response generation unit, configured to generate understanding result information of the speech information using a target language model corresponding to the text expression type;

[0016] The understanding output unit is used to output the understanding result information through the vehicle-mounted terminal.

[0017] According to a third aspect of the present disclosure, there is provided an electronic device, including:

[0018] at least one processor; and

[0019] a memory communicatively connected to the at least one processor; wherein,

[0020] The memory stores instructions that can be executed by the at least one processor. The instructions are executed by the at least one processor to enable the at least one processor to perform the method according to the first aspect.

[0021] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute the method according to the first aspect.

[0022] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising: a computer program, wherein the computer program is stored in a readable storage medium, at least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program so that the electronic device executes the method of the first aspect.

[0023] The technology disclosed in the present invention solves the technical problem that the vehicle computer can only handle simple voice interaction scenarios. By acquiring the voice information corresponding to the target user collected by the vehicle terminal device, and acquiring the text information corresponding to the voice information. The text expression type corresponding to the voice information can be determined through the text information, and the understanding of the voice information can be achieved by adopting the target language model corresponding to the text expression type, and the corresponding understanding result information can be obtained. By associating the target language model with the text expression type of the voice information, the target language model involved in the understanding of the voice information can be made more compatible with the voice information, and the understanding result information with higher accuracy can be quickly obtained. By outputting the understanding result information through the vehicle terminal, fast and accurate interaction with the target user can be achieved.

[0024] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0026] FIG1 is a schematic diagram of a first embodiment of the present disclosure;

[0027] FIG2 is a schematic diagram of a second embodiment of the present disclosure;

[0028] FIG3 is a schematic diagram of a third embodiment of the present disclosure;

[0029] FIG4 is a schematic flow chart of speech understanding according to an embodiment of the present disclosure;

[0030] FIG5 is a schematic diagram of a fourth embodiment according to the present disclosure;

[0031] FIG6 is a schematic diagram of a fifth embodiment according to the present disclosure;

[0032] FIG7 is a schematic diagram of an application of a voice interaction method according to an embodiment of the present disclosure;

[0033] FIG8 is a schematic diagram according to a sixth embodiment of the present disclosure;

[0034] FIG9 is a block diagram of an electronic device for implementing the voice interaction method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0035] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0036] The present disclosure provides a voice interaction method, device, equipment, medium and product, which are applied to technical fields such as autonomous driving, large language models and smart cockpits in the field of artificial intelligence to achieve the purpose of quickly and accurately realizing vehicle-computer interaction.

[0037] In related technologies, the car computer can be configured in the vehicle or the cockpit to achieve assisted driving through human-computer interaction. In the more common assisted driving scenarios, the car computer can collect the user's corresponding voice information, determine the corresponding text understanding model based on the voice information, and understand the actual meaning of the voice information through the text understanding model. And by understanding the meaning of the voice information, corresponding human-computer interaction is carried out. For example, the voice of the user initiating "navigation to place A" can be understood through the NLU (Natural Language Understanding, natural language processing) model voice, that is, the two key words navigation and place A. However, at present, the NLU model can only understand simple voice content, and the understanding accuracy is not high. As a result, in actual applications, users can only send corresponding voices through fixed texts, which makes the car-computer interaction scenario more limited.

[0038] To address the above issues, this disclosure considers using large language models to understand speech content. However, large language models have a large number of parameters. While they can guarantee accurate understanding, they also reduce speed and efficiency. Continuing to use existing NLU models, while they maintain recognition speed and efficiency, reduces accuracy. Therefore, this disclosure considers processing based on the text classification of user voice messages.

[0039] Therefore, in the technical solution disclosed in the present invention, it is possible to obtain the voice information corresponding to the target user collected by the vehicle-mounted terminal device, and obtain the text information corresponding to the voice information. The text expression type corresponding to the voice information can be determined through the text information, and the understanding of the voice information can be achieved by adopting the target language model corresponding to the text expression type, and the corresponding understanding result information can be obtained. By associating the target language model with the text expression type of the voice information, the target language model involved in the understanding of the voice information can be made more compatible with the voice information, and understanding result information with higher accuracy can be quickly obtained. By outputting the understanding result information through the vehicle-mounted terminal, fast and accurate interaction with the target user can be achieved.

[0040] The technical solution of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0041] FIG1 is a schematic diagram of a first embodiment of the present disclosure, and is a schematic diagram of a system for implementing a voice interaction method as shown in FIG1 .

[0042] In the system, the vehicle-mounted terminal 11 can interact with the target user and collect the target user's voice information. The vehicle-mounted terminal 11 can also establish a communication connection with the server 12 and send the voice information to the server 12.

[0043] The server 12 can receive the voice information sent by the vehicle-mounted terminal device 11 and obtain the text information corresponding to the voice information. Based on the text information, the server 12 determines the corresponding text expression type. Using the target language model corresponding to the text expression type, the server 12 determines the understanding result of the voice information. The server 12 outputs the understanding result information via the vehicle-mounted terminal, thus achieving human-computer interaction.

[0044] FIG2 is a schematic diagram of a second embodiment of the present disclosure. Referring to the voice interaction method shown in FIG2 , the following steps may be included:

[0045] 201. Acquire voice information corresponding to a target user collected by an in-vehicle terminal, and acquire text information corresponding to the voice information.

[0046] Optionally, the vehicle-mounted terminal may refer to a device configured in a vehicle or a cabin, such as a vehicle computer, a tablet computer, etc. In this embodiment, there are no excessive restrictions on the specific type of the vehicle-mounted terminal.

[0047] The voice interaction method disclosed herein can be configured in an in-vehicle terminal or in an electronic device that is communicatively connected to the in-vehicle terminal, such as a server, a computer, or the like. The server can be, for example, a common server or a cloud server, and this embodiment does not impose any particular restrictions on the type of electronic device.

[0048] Wherein, obtaining the voice information corresponding to the target user collected by the vehicle-mounted terminal in step 201 may include: receiving the voice information corresponding to the target user collected by the vehicle-mounted terminal device. The voice information may refer to a voice signal collected by a microphone.

[0049] Acquiring text information corresponding to the voice information may include: recognizing the text information in the voice information based on a voice recognition model. The voice recognition model may convert the voice information into text information.

[0050] Text information may refer to information consisting of multiple characters.

[0051] 202. Determine a text expression type corresponding to the voice information based on the text information.

[0052] The text expression type may refer to a specific expression method of the text information, and may be an imperative expression type or a non-imperative expression type.

[0053] Step 202 may include determining a text expression type corresponding to the speech information based on the complexity of the text information. Classifying text expression types based on the complexity of the text information can reflect the complexity of the text information. The target language model determined based on the text expression type is then matched to the complexity of the text information, thereby achieving a balance between recognition accuracy and efficiency.

[0054] 203. Use a target language model corresponding to the text expression type to determine understanding result information of the speech information.

[0055] The target language model may refer to a natural language processing (NLU) model used to implement text understanding. A natural language processing model can understand and analyze specific content in natural language text or speech through a computer.

[0056] The understanding result information may refer to information formed by the meaning of the text information understood by the NLU model.

[0057] Optionally, step 203 may include: inputting text information of the speech information into a target language model to obtain understanding result information of the speech information.

[0058] Furthermore, understanding result information may refer to information represented by three dimensions: domain, intent, and slot value.

[0059] For example, if the voice message is "Book me a ticket to Beijing", the understanding result information may refer to the domain: air tickets, the intention: book air tickets, and the slot value: destination = Beijing. The understanding result information can accurately represent the understanding content of the voice message.

[0060] 204. Output the understanding result information through the vehicle terminal.

[0061] Optionally, step 204 may include: determining an application corresponding to the understanding result information, and implementing a corresponding interactive operation by calling the application of the vehicle-mounted terminal.

[0062] Taking the above-mentioned "Help me book a ticket to Beijing" as an example, you can determine that you need to book a ticket to Beijing by understanding the result information, so you can call the ticket booking program, start the ticket booking program of the on-board terminal, and control the ticket booking program of the on-board terminal to use Beijing as the destination as the query target, query the remaining tickets, display the remaining tickets, and complete the ticket booking through interaction with the user.

[0063] In the technical solution disclosed in the present invention, the voice information corresponding to the target user collected by the vehicle-mounted terminal device can be obtained, and the text information corresponding to the voice information can be obtained. The text expression type corresponding to the voice information can be determined through the text information, and the understanding of the voice information can be achieved by adopting the target language model corresponding to the text expression type, and the corresponding understanding result information can be obtained. By associating the target language model with the text expression type of the voice information, the target language model involved in the understanding of the voice information can be made more compatible with the voice information, and the understanding result information with higher accuracy can be quickly obtained. By outputting the understanding result information through the vehicle-mounted terminal, fast and accurate interaction with the target user can be achieved.

[0064] In order to enable readers to have a deeper understanding of the implementation principle of the present disclosure, the embodiment shown in FIG2 is further refined in combination with the following FIG3 to FIG7 .

[0065] In practical applications, more than two language processing models can be provided to achieve accurate transmission of the terminal device according to the text expression type corresponding to the voice information.

[0066] Therefore, as an embodiment, a target language model corresponding to the text expression type is used to generate understanding result information of the speech information, including:

[0067] A target language model corresponding to the text expression type is determined from at least two preset language processing models, where the at least two language processing models include a text processing model and a large language model, and the number of parameters of the text processing model is smaller than the number of parameters of the large language model.

[0068] The target language model is used to determine the understanding result of the speech information.

[0069] The text processing model may be a type of language processing model, and may refer to a small language processing model with fewer parameters.

[0070] Large Language Model (LLM) A large language model can be a type of NLU, specifically a deep learning model that is trained using a large amount of text data and has a relatively large number of parameters.

[0071] Large language models are larger than text processing models. They have more model parameters and can understand and process more complex text.

[0072] Optionally, the language processing model may include an input layer, a hidden layer, and an output layer.

[0073] The input layer of the text processing model and the input layer of the large language model can be different. Compared with the input layer of the text processing model, the input layer of the large language model can be more complex, including more input neurons, to receive more complex text information.

[0074] The hidden layers of the text processing model and the large language model can be different. Compared to the hidden layers of the text processing model, the hidden layers of the large language model can be more complex, containing more hidden neurons, to extract features from more complex text information and obtain more accurate text features.

[0075] In addition, the output layer of the text processing model and the output layer of the large language model can be different. Compared with the output layer of the text processing model, the output layer of the large language model is also more complex. The output layer of the text processing model can understand and process the text of the imperative expression type. The large language model can understand and process the text features of the non-imperative expression type to obtain understanding result information. The non-imperative expression type can include at least one, that is, the output layer of the large language model can understand the text features of at least one non-imperative expression type. Then the output layer of the large language model can obtain the understanding result information corresponding to the non-imperative expression type corresponding to the text feature in the at least one non-imperative expression type.

[0076] The technical solution disclosed in the present invention determines the target language model corresponding to the text expression type from at least two preset language processing models, so that the target language model is a language processing model corresponding to the text expression type, so that the target language model is not limited to a single language processing model and is applicable to more text understanding scenarios. Furthermore, when the understanding result information of the speech information is obtained through the target language model, both processing efficiency and processing accuracy can be taken into account, thereby effectively improving the speech understanding effect.

[0077] FIG3 is a schematic diagram of a third embodiment of the present disclosure. Referring to the voice interaction method shown in FIG3 , the method differs from the previous embodiments in that determining a target language model corresponding to a text expression type from at least two preset language processing models may include:

[0078] 301. Determine whether the text expression type is an imperative expression type. If so, execute 302; if not, execute 304.

[0079] The imperative expression type refers to an expression type expressed according to a command expression structure, and the command expression structure includes: a verb and an action target corresponding to the verb.

[0080] The imperative expression type can directly extract the verb and the action target corresponding to the verb from the command expression result. The verb in the command expression structure can be used to determine the intent, and the action target corresponding to the verb can be used to determine the slot value.

[0081] For example, "Navigate to location A" can be an imperative expression type. From "Navigate to location A", the understanding content can be extracted as follows: Intent: Navigation, Slot value: Target location = location A.

[0082] 302. Determine a text processing model as a target language model from at least two preset language processing models.

[0083] 303. Using the text processing model, determine the understanding result information of the speech information.

[0084] Optionally, using the text processing model to determine the understanding result information of the speech information may include: inputting text information of the speech information into the text processing model, and obtaining the understanding result information output by the text processing model.

[0085] 304. Determine whether the text expression type is a non-imperative expression type. If so, execute 304.

[0086] Optionally, in this embodiment, a type identifier may be set for the text expression type. For example, an imperative expression type may be set to a first type identifier, and a non-imperative expression type may be set to a second type identifier.

[0087] If the text expression type is a first type identifier, the text expression type is determined to be an imperative expression type. If the text expression type is a second type identifier, the text expression type is determined to be a non-imperative expression type.

[0088] Furthermore, step 301 may include: determining whether the type identifier of the text expression type is a first type identifier of an imperative expression type. Step 304 may include: determining whether the type identifier of the text expression type is a second type identifier of a non-imperative expression type.

[0089] 305. Determine a large language model as a target language model from at least two preset language processing models.

[0090] Optionally, the text expression complexity of the non-imperative expression type is greater than the text expression complexity of the imperative expression type.

[0091] Using a large language model as the target language model can achieve more complex understanding of text information.

[0092] 306. Using a large language model, determine the understanding result information of the speech information.

[0093] Optionally, step 306 may include: sending text information of the voice information to a large language model to obtain understanding result information output by the large language model.

[0094] In practical applications, when using a large language model to determine the understanding result information of speech information, the background speech information of the speech information can be obtained first, and the understanding result information in the speech information and the background speech information can be identified through the large language model.

[0095] In the technical solution disclosed in the present invention, the text expression type of the voice information can be any one of the imperative expression type and the non-imperative expression type, and the expression complexity of the non-imperative expression type is greater than the expression complexity of the imperative expression type. When the text expression type is the imperative expression type, it means that the text information expression of the voice information is relatively concise and clear, and the understanding of the text information can be achieved through an ordinary text processing model, thereby improving the efficiency of text understanding. When the text expression type is the non-imperative expression type, it means that the text information expression of the voice information is relatively complex, and the understanding of the text information can be completed through a large language model that can recognize complex texts, thereby improving the accuracy of text understanding. Therefore, by matching the text expression type to the corresponding target language model, the understanding of the voice information can be achieved in both speed and accuracy dimensions.

[0096] Furthermore, based on any of the above embodiments, after determining that the text processing model is the target language model, the method further includes:

[0097] If the understanding result information of the speech information is not obtained through the text processing model, the large language model is determined to be the target language model.

[0098] In order to facilitate understanding of the technical solution of the present disclosure, FIG4 shows a schematic flow chart of speech understanding. Referring to FIG4, after obtaining the speech information, the text information in the speech information can be obtained, and then the text expression type corresponding to the text information can be determined. Determine whether the text expression type is an imperative expression type. If so, the text processing model can be determined to be the target language model. If not, the large language model can be determined to be the target language model. The text information is understood by the text processing model, and it is determined whether the text processing model obtains the understanding result information. If so, the understanding result information is determined. Otherwise, the large language model is determined to be the target language model.

[0099] Furthermore, when the large language model is the target language model, understanding result information determined by the large language model is obtained.

[0100] In the technical solution disclosed herein, when the text processing model fails to obtain the speech information's understanding results, the large language model can be used as the target language model. By using the large language model to reacquire the speech information's understanding results, this avoids gaps in understanding caused by the text processing model's comprehension failures, effectively improving the success rate of speech information comprehension.

[0101] Furthermore, based on any of the above embodiments, determining the text expression type corresponding to the voice information according to the text information includes:

[0102] Based on the text classifier obtained through training, the text expression of the text information is classified to obtain the text expression type corresponding to the voice information. The text classifier includes an imperative expression type and a non-imperative expression type.

[0103] Optionally, the imperative expression type and the non-imperative expression type can be used as two classification categories of the text classifier. Classifying the text expression of the text information based on the trained text classifier to obtain the text expression type corresponding to the speech information can include: inputting the text information into the trained text classifier to obtain the text expression type corresponding to the speech information.

[0104] The text expression type obtained by the text classifier may be an imperative expression type or a non-imperative expression type.

[0105] In the technical solution disclosed herein, the text expression of text information is classified by a text classifier, which can realize automatic classification of the text expression of voice information and improve the classification efficiency and accuracy of text information.

[0106] Optionally, the non-imperative expression type includes at least one of a natural expression type, a referential expression type, an omitted expression type, a grammatical error type, and a pause expression type.

[0107] Natural expression type refers to the expression type in which text is expressed using natural language expressions.

[0108] Referential expression type refers to the expression type in which pronouns are used to refer to entity words in the text.

[0109] The grammatical error type refers to the type of expression that contains grammatical errors and / or expression errors in the text.

[0110] The omitted expression type refers to the expression type in which the key entity words are omitted in the text.

[0111] The pause expression type refers to the expression type in which there is an expression connection between the previous and next texts.

[0112] Alternatively, natural language expressions can be more colloquial, lacking command structures. For example, "I've been busy since getting up today and haven't had dinner yet. Any recommendations for places?" It's difficult to directly extract short sentences with command structures from this text. A large language model capable of recognizing more complex text information is needed to understand more colloquial text and obtain more accurate understanding results.

[0113] The expression method that uses pronouns to refer to entity words in the referential expression type means that pronouns appear in the text information. The pronouns can be, for example, any one of personal pronouns, possessive pronouns, demonstrative pronouns, reflexive pronouns, reciprocal pronouns, interrogative pronouns, relative pronouns, conjunctive pronouns, indefinite pronouns, substitute words, etc. The pronouns can be, for example, "she, he, it", "this, which, that", etc. For example, sentence 1 is "Introduce scenic spot T", and sentence 2 is "Sounds good. Navigate there." When sentence 2 is understood as text information, it is necessary to determine that the location corresponding to "there" is "scenic spot T". It is necessary to obtain the background sentence of sentence 2, that is, sentence 1, and perform language recognition processing. A large language model can be used to confirm the reference of the text information of the referential expression type for semantic understanding and obtain accurate understanding result information.

[0114] There are grammar errors and / or expression errors in the text of the grammar error type. For example, it can include phenomena such as repeated words and incorrect word usage. For example: There are repeated characters "tian" in "Navigate to tian tian tian tian di". The word "bo" in "Dial S's phone number" is incorrect. A large language model can be used to correct the text information of the grammar error type for semantic understanding and obtain accurate understanding result information.

[0115] The key entity words omitted in the ellipsis expression type can be words related to the subject, predicate, object, attributive, adverbial or complement, etc. in the text. The key entity words can be words that play a key role in text understanding. For example, sentence 1 is "Introduce scenic spot T", and sentence 2 is "Sounds like the T1 location of scenic spot T is good. Hurry up and navigate there." The complement is lacking for the verb "navigate" in sentence 2. According to the context, it can be determined that "the T1 location of scenic spot T" is the complement of the verb "navigate". A large language model can be used to complete the words of the text information of the ellipsis expression type for semantic understanding and obtain accurate understanding result information.

[0116] The existence of a statement connection between the front and back texts in the pause expression type can mean that the two texts spliced together are a complete sentence. For example, sentence 1 is "Navigate to city D", and sentence 2 is "the museum". When sentence 2 is used as a voice signal, it is not a complete sentence and can be determined as the pause expression type. A large language model can be used to splice the pause expression type and the previous sentence to perform semantic understanding on the spliced complete sentence and obtain accurate understanding result information.

[0117] Of course, in practical applications, the non-imperative expression type can also include other expression types. The natural expression type, referential expression type, ellipsis expression type, grammar error type, and pause expression type listed in this embodiment are only exemplary and should not constitute a specific limitation on the non-imperative expression type.

[0118] In the technical solution disclosed herein, at least one of the natural expression type, referential expression type, elliptical expression type, grammatical error type, and pause expression type is defined as a non-imperative expression type. This allows the non-imperative expression type to include at least one text expression type, enabling more complex text classification, improving the classification effectiveness of text expression types, and avoiding text expression type classification issues caused by the single composition of non-imperative expression types.

[0119] FIG5 is a schematic diagram of a fourth embodiment of the present disclosure. Referring to the text classifier training method shown in FIG5 , the method may include:

[0120] 501. Collect first training data according to an imperative expression type. The first training data includes training text expressed according to a command expression structure corresponding to the imperative expression type and a type label corresponding to the imperative expression type.

[0121] 502. Collect second training data according to the non-imperative expression type. The second training data includes training text collected according to the non-imperative expression type and a type label corresponding to the non-imperative expression type.

[0122] 503. Train and obtain a text classifier based on the first training data and the second training data.

[0123] Optionally, collecting the first training data according to the imperative expression type may include: reading multiple candidate texts from a corpus, selecting a first candidate text represented by a command expression structure from the multiple candidate texts, and obtaining the first training data corresponding to the imperative expression type.

[0124] Optionally, collecting the second training data according to the non-imperative expression type may include reading multiple candidate texts from a corpus, selecting a second candidate text represented by the non-imperative expression structure from the multiple candidate texts, and obtaining the second training data corresponding to the non-imperative expression type.

[0125] Of course, in actual applications, after obtaining the first and second candidate texts, manual verification can be performed to determine whether the first candidate text belongs to the imperative expression type, and whether the second candidate text belongs to the non-imperative expression type. Manual verification can ensure that the training text corresponds to the training requirements, thereby making the text classifier training process more efficient.

[0126] In the technical solution disclosed herein, a text classifier is trained using first training data of an imperative expression type and second training data of a non-imperative expression type. This allows the text classifier to learn both training text expressed using imperative expression results and training text collected using non-imperative expression types, effectively classifying both types of text and achieving higher classification accuracy.

[0127] FIG6 is a schematic diagram of a fifth embodiment of the present disclosure. Referring to the schematic diagram shown in FIG6 , the difference from the previous embodiment is that, based on the text information, determining the text expression type corresponding to the voice information may include:

[0128] 601. Determine text complexity corresponding to the text information.

[0129] 602. If the text complexity is less than or equal to the complexity threshold, determine that the imperative expression type is the text expression type corresponding to the voice information.

[0130] 603. If the text complexity is greater than the complexity threshold, determine that the non-command expression type is the text expression type corresponding to the voice information.

[0131] Optionally, text complexity may refer to the complexity of text information, and specifically may be a complexity probability value.

[0132] Step 601 may include determining the text complexity corresponding to the text information using a text complexity prediction model. The text complexity prediction model may be a trained deep learning model for predicting text complexity. The text complexity prediction model may be trained.

[0133] Specifically, multiple pieces of complex training data can be obtained, each piece of which can correspond to a real complexity. A text complexity prediction model can be trained based on the multiple pieces of complex training data and the real complexity of each piece of complex training data. The input of the text complexity prediction model is the text to be predicted, and the output is the complexity of the text to be predicted.

[0134] The complexity threshold may be a constant set according to usage requirements. The complexity threshold may be greater than 0 and less than 1. For example, the complexity threshold may be 0.5.

[0135] Wherein, step 601 may further specifically include: using at least one complex factor, performing probability analysis on the text information in each complex factor, and obtaining the complex probability corresponding to the text information in the at least one complex factor.

[0136] Furthermore, the complexity probabilities corresponding to the at least one complexity factor can be weighted and summed to obtain the text complexity of the text information. In addition, the complexity probability corresponding to the at least one complexity factor can be directly determined as the text complexity.

[0137] Optionally, the weights corresponding to the at least one complexity factor may be the same. Of course, the weights corresponding to the at least one complexity factor may also be different. The weights corresponding to the at least one complexity factor may be set according to usage requirements.

[0138] The at least one complexity factor may include a complexity factor corresponding to at least one text expression type. Specifically, the at least one complexity factor may include at least one of an expression complexity factor corresponding to a natural expression type, a reference complexity factor corresponding to a referential expression type, a grammatical complexity factor corresponding to a grammatical error type, an omission complexity factor corresponding to an omission expression type, and a pause complexity factor corresponding to a pause expression type.

[0139] Specifically, the text information can be segmented to obtain at least one word and the word type corresponding to at least one word; according to the word type corresponding to at least one word, the probability value corresponding to at least one complex factor is obtained to obtain the complex probability corresponding to at least one complex factor.

[0140] Furthermore, for any complex factor, the number of words or word types associated with the complex factor in at least one word of the text information and the word type corresponding to at least one word can be determined to obtain the text-related quantity of the complex factor. Then, the text-related quantity corresponding to the at least one complex factor is obtained, and then the text-related quantity corresponding to the at least one complex factor is normalized to obtain the complex probability corresponding to the at least one complex factor. Of course, the method for obtaining the complex probability of the complex factor shown in this embodiment is only exemplary and should not constitute a detailed limitation of the technical solution disclosed herein.

[0141] In this embodiment, if the text complexity is obtained by weighted summing the complexity probabilities corresponding to at least one complexity factor, the value obtained by weighted summing is the text complexity of the obtained text information. In this case, the text expression type can be determined by directly comparing the complexity threshold with the text complexity.

[0142] In this embodiment, if the text complexity is composed of the complexity probabilities corresponding to at least one complexity factor, then the complexity threshold may be the factor threshold corresponding to at least one complexity factor.

[0143] Furthermore, if the complexity probability of one or more complex factors is greater than the corresponding factor threshold, then the non-imperative expression type is determined to be the text expression type corresponding to the voice information. If the complexity probability of all complex factors in at least one complex factor is less than the corresponding factor threshold, then the imperative expression type is determined to be the text expression type corresponding to the voice information.

[0144] In the technical solution disclosed in the present invention, the text expression type of the text information is determined based on the relationship between the text complexity of the text information and the complexity threshold. If the text complexity is less than or equal to the complexity threshold, the imperative expression type is determined to be the text expression type corresponding to the voice information. If the text complexity is greater than the complexity threshold, the non-imperative expression type is determined to be the text expression type corresponding to the voice information. By confirming the text expression type of the text information through complexity, the determination of the target language model can be achieved through text complexity. The text information with higher complexity is then processed using a large language model to ensure the accuracy of text understanding. The text information with lower complexity is processed using a text processing model to ensure the efficiency of text understanding.

[0145] Furthermore, based on any of the above embodiments, outputting understanding result information through the vehicle-mounted terminal includes:

[0146] A target application corresponding to the understanding result information is determined from at least one application supported by the in-vehicle terminal.

[0147] Generate interactive results corresponding to the understanding result information based on the target application.

[0148] The interaction results are executed through the target application of the vehicle terminal.

[0149] Optionally, the at least one application supported by the in-vehicle terminal may refer to an application that can be run on the in-vehicle terminal.

[0150] Determining a target application corresponding to the understanding result information from at least one application supported by the vehicle terminal may include: obtaining a program function corresponding to each of the at least one application supported by the vehicle terminal, obtaining a field in the understanding result information, and determining a target program function that matches the field in the understanding result information from the program functions corresponding to the at least one application. Furthermore, determining the application corresponding to the target program function as the target application.

[0151] Furthermore, the target program function that matches the field in the understanding result information may specifically refer to the target program function including the field in the understanding result information.

[0152] For example, taking the example of at least one application including an electronic map application, a music player application, and a phone call application, and assuming that the domain of the understanding result information is the navigation domain, the electronic map application with navigation function is determined to be the target application.

[0153] Optionally, generating an interaction result corresponding to the understanding result information based on the target application may include obtaining the intent and slot value in the understanding result information, generating an operation instruction for the target application based on the intent and slot value, and sending the operation instruction to the calling interface of the target application of the vehicle-mounted terminal device. The operation instruction instructs the target application to perform corresponding operations according to the intent and slot value.

[0154] In the technical solution disclosed herein, after obtaining understanding result information, a target application corresponding to the understanding result information can be determined from at least one application supported by the vehicle terminal. By first determining the target application corresponding to the understanding result information, the vehicle terminal's interactive object can be acquired. After generating an interaction result corresponding to the understanding result information based on the target application, the interaction result can be executed by the vehicle terminal's target application, completing the interaction from the understanding result information to the vehicle terminal and improving the user's interactive experience.

[0155] In order to facilitate understanding of the technical solution of the present disclosure, FIG7 shows an application diagram of the voice interaction method of the present disclosure. Referring to FIG7 , it is assumed that the target user 701 inputs two voices, namely voice 1 "Please introduce H Mountain" and voice 2 "It sounds good, please navigate there". Voice 2 may be the latest voice information corresponding to the target user 701. Inputting voice 2 into the text classifier can determine that the text information of voice 2 is a referential expression type. It is then determined that voice 2 is a non-imperative expression type, so the large language model can be determined as the target language model of voice 2. The large language model can then be used to determine the understanding result information 702 corresponding to voice 2 and voice 1 associated with voice 2.

[0156] Target user 703 inputs speech 3, "Navigate to location B." Inputting speech 3 into the text classifier determines that the text information in speech 3 is an imperative expression type. Therefore, the text processing model is determined to be the target language model for speech 3. Furthermore, the text processing model can be used to determine the understanding result information 704 of speech 3.

[0157] Figure 8 is a schematic diagram of the sixth embodiment of the present disclosure. Referring to the voice interaction device shown in Figure 8, it may include: an information acquisition unit 801, a type determination unit 802, a response generation unit 803 and an understanding output unit 804.

[0158] The information acquisition unit 801 is used to obtain voice information corresponding to the target user, collected by the vehicle terminal, and obtain the text information corresponding to the voice information. The type determination unit 802 is used to determine the text expression type corresponding to the voice information based on the text information. The response generation unit 803 is used to generate the understanding result information of the voice information using the target language model corresponding to the text expression type. The understanding output unit 804 is used to output the understanding result information via the vehicle terminal.

[0159] Furthermore, based on any of the above embodiments, the reply generating unit includes:

[0160] A model selection module is used to determine a target language model corresponding to the text expression type from at least two preset language processing models, where the at least two language processing models include a text processing model and a large language model, and the number of parameters of the text processing model is smaller than the number of parameters of the large language model.

[0161] The information understanding module is used to determine the understanding result information of the speech information through the target language model.

[0162] Furthermore, based on any of the above embodiments, the model selection module includes:

[0163] The first selection submodule is used to determine the text processing model as the target language model from at least two preset language processing models if the text expression type is an imperative expression type. The imperative expression type refers to an expression type expressed according to a command expression structure, and the command expression structure includes: a verb and an action target corresponding to the verb.

[0164] Furthermore, based on any of the above embodiments, after determining that the text processing model is the target language model, the method further includes:

[0165] If the understanding result information of the speech information is not obtained through the text processing model, the large language model is determined to be the target language model.

[0166] Furthermore, based on any of the above embodiments, the model selection module includes:

[0167] The second selection submodule is used to determine a large language model as a target language model from at least two preset language processing models if the text expression type is a non-imperative expression type, and the expression complexity of the non-imperative expression type is greater than the expression complexity of the imperative expression type.

[0168] Furthermore, based on any of the above embodiments, the type determination unit includes:

[0169] The text classification unit is used to classify the text expression of text information based on the text classifier obtained through training, and obtain the text expression type corresponding to the voice information. The text classifier includes imperative expression type and non-imperative expression type.

[0170] Furthermore, based on any of the above embodiments, the non-imperative expression type includes at least one of: a natural expression type, a referential expression type, an omitted expression type, a grammatical error type, and a pause expression type.

[0171] Natural expression type refers to the expression type in which text is expressed using natural language expressions.

[0172] Referential expression type refers to the expression type in which pronouns are used to refer to entity words in the text.

[0173] The grammatical error type refers to the type of expression that contains grammatical errors and / or expression errors in the text.

[0174] The omitted expression type refers to the expression type in which the key entity words are omitted in the text.

[0175] The pause expression type refers to the expression type in which there is an expression connection between the previous and next texts.

[0176] Furthermore, based on any of the above embodiments, the method further includes:

[0177] The first collection unit is used to collect first training data according to the imperative expression type, where the first training data includes training text expressed according to the command expression structure corresponding to the imperative expression type and a type label corresponding to the imperative expression type.

[0178] The second collecting unit is used to collect second training data according to the non-imperative expression type, where the second training data includes training text collected according to the non-imperative expression type and a type label corresponding to the non-imperative expression type.

[0179] The model training unit is used to train a text classifier based on the first training data and the second training data.

[0180] Furthermore, based on any of the above embodiments, the type determination unit includes:

[0181] The complexity determination module is used to determine the text complexity corresponding to the text information.

[0182] The first determination module is configured to determine that the imperative expression type is the text expression type corresponding to the voice information if the text complexity is less than or equal to the complexity threshold.

[0183] The second determination module is configured to determine that the non-imperative expression type is the text expression type corresponding to the voice information if the text complexity is greater than the complexity threshold.

[0184] Furthermore, based on any of the above embodiments, the output unit is understood to include:

[0185] The program determination module is used to determine a target application corresponding to the understanding result information from at least one application supported by the vehicle terminal.

[0186] The result generation module is used to generate interactive results corresponding to the understanding result information according to the target application.

[0187] The result output module is used to execute the interaction results through the target application of the vehicle terminal.

[0188] It should be noted that the at least two language processing models in this embodiment are not language processing models for a specific user and cannot reflect the personal information of a specific user. It should be noted that the training data in this embodiment comes from a public data set.

[0189] In the technical solutions disclosed herein, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0190] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0191] According to an embodiment of the present disclosure, the present disclosure also provides a computer program product, which includes: a computer program, the computer program is stored in a readable storage medium, at least one processor of an electronic device can read the computer program from the readable storage medium, and at least one processor executes the computer program so that the electronic device executes the solution provided by any of the above embodiments.

[0192] FIG9 shows a schematic block diagram of an example electronic device 900 that can be used to implement an embodiment of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0193] As shown in Figure 9, the device 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. Various programs and data required for the operation of the device 900 can also be stored in the RAM 903. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0194] Various components in the device 900 are connected to the I / O interface 905, including an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a magnetic disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0195] The computing unit 901 can be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 901 performs the various methods and processes described above, such as the voice interaction method. For example, in some embodiments, the voice interaction method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the voice interaction method described above can be performed. Alternatively, in other embodiments, the computing unit 901 can be configured to perform the voice interaction method in any other appropriate manner (e.g., by means of firmware).

[0196] Various embodiments of the systems and techniques described above can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0197] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. Such program code can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0198] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0199] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0200] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0201] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact via a communication network. This client-server relationship is established by computer programs running on the respective computers, establishing a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host, a host product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosts and VPS services ("Virtual Private Servers" or simply "VPS"). The server may also be a server in a distributed system or a server integrated with blockchain.

[0202] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not a limitation herein.

[0203] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A voice interaction method, comprising: Acquire voice information corresponding to the target user collected by the vehicle-mounted terminal, and acquire text information corresponding to the voice information; Determining, according to the text information, a text expression type corresponding to the voice information; Using a target language model corresponding to the text expression type, determining understanding result information of the speech information; The understanding result information is outputted through the vehicle-mounted terminal.

2. The method according to claim 1, wherein: The step of using a target language model corresponding to the text expression type to generate understanding result information of the speech information includes: Determining a target language model corresponding to the text expression type from at least two preset language processing models, wherein the at least two language processing models include a text processing model and a large language model, and the number of parameters of the text processing model is smaller than the number of parameters of the large language model; The understanding result information of the speech information is determined through the target language model.

3. The method according to claim 2, wherein: The step of determining a target language model corresponding to the text expression type from at least two preset language processing models includes: If the text expression type is an imperative expression type, the text processing model is determined as the target language model from at least two preset language processing models, and the imperative expression type refers to an expression type expressed according to a command expression structure, and the command expression structure includes: a verb and an action target corresponding to the verb.

4. The method according to claim 3, wherein: After determining that the text processing model is the target language model, the method further includes: If the understanding result information of the speech information is not obtained through the text processing model, the large language model is determined to be the target language model.

5. The method according to any one of claims 2 to 4, wherein: The step of determining a target language model corresponding to the text expression type from at least two preset language processing models includes: If the text expression type is a non-imperative expression type, the large language model is determined as the target language model from among at least two preset language processing models.

6. The method according to any one of claims 1 to 5, wherein: The determining, according to the text information, a text expression type corresponding to the voice information includes: Based on the text classifier obtained through training, the text expression of the text information is classified to obtain The text expression type corresponding to the voice information, the text classifier includes an imperative expression type and a non-imperative expression type.

7. The method according to claim 6, wherein: The non-imperative expression type includes: at least one of a natural expression type, a referential expression type, an omitted expression type, a grammatical error type, and a pause expression type; The natural expression type refers to the expression type in which the text is expressed using natural language expression; The referential expression type refers to the expression type in which a pronoun is used to refer to an entity word in a text; The grammatical error type refers to the type of expression in which there are grammatical errors and / or expression errors in the text; The omitted expression type refers to an expression type in which key entity words are omitted in the text; The pause expression type refers to an expression type in which there is an expression association between the preceding and following texts.

8. The method according to claim 6, wherein: Also includes: According to the imperative expression type, first training data is collected, where the first training data includes training text expressed according to the command expression structure corresponding to the imperative expression type and a type label corresponding to the imperative expression type; According to the non-imperative expression type, second training data is collected, where the second training data includes training text collected according to the non-imperative expression type and a type label corresponding to the non-imperative expression type; The text classifier is trained based on the first training data and the second training data.

9. The method according to any one of claims 1 to 8, wherein: The determining, according to the text information, a text expression type corresponding to the voice information includes: Determining the text complexity corresponding to the text information; If the text complexity is less than or equal to the complexity threshold, determining the imperative expression type as the text expression type corresponding to the voice information; If the text complexity is greater than the complexity threshold, the non-imperative expression type is determined to be the text expression type corresponding to the voice information.

10. The method according to any one of claims 1 to 9, wherein: The outputting the understanding result information through the vehicle-mounted terminal includes: Determine a target application corresponding to the understanding result information from at least one application supported by the vehicle-mounted terminal; generating, according to the target application, an interaction result corresponding to the understanding result information; The interaction result is played through the target application of the vehicle terminal.

11. A voice interaction device, comprising: The information acquisition unit is used to acquire the voice information corresponding to the target user collected by the vehicle terminal, and obtain the Text information corresponding to the voice information; A type determination unit, configured to determine a text expression type corresponding to the voice information according to the text information; A response generation unit, configured to generate understanding result information of the speech information by using a target language model corresponding to the text expression type; The understanding output unit is used to output the understanding result information through the vehicle-mounted terminal.

12. The device according to claim 11, wherein The reply generating unit comprises: A model selection module, used to determine a target language model corresponding to the text expression type from at least two preset language processing models, wherein the at least two language processing models include a text processing model and a large language model, and the number of parameters of the text processing model is smaller than the number of parameters of the large language model; The information understanding module is used to determine the understanding result information of the speech information through the target language model.

13. The device according to claim 12, wherein: The model selection module comprises: The first selection submodule is used to determine the text processing model as the target language model from at least two preset language processing models if the text expression type is an imperative expression type, and the imperative expression type refers to an expression type expressed according to a command expression structure, and the command expression structure includes: a verb and an action target corresponding to the verb.

14. The device according to claim 13, wherein: After determining that the text processing model is the target language model, the method further includes: If the understanding result information of the speech information is not obtained through the text processing model, the large language model is determined to be the target language model.

15. The device according to any one of claims 11 to 14, wherein: The model selection module comprises: The second selection submodule is used to determine the large language model as the target language model from at least two preset language processing models if the text expression type is a non-imperative expression type, and the expression complexity of the non-imperative expression type is greater than the expression complexity of the imperative expression type.

16. The device according to any one of claims 11 to 15, wherein: The type determination unit comprises: The text classification unit is used to classify the text expression of the text information based on the text classifier obtained through training, and obtain the text expression type corresponding to the voice information, wherein the text classifier includes an imperative expression type and a non-imperative expression type.

17. The device according to claim 16, wherein: The non-imperative expression type includes: at least one of a natural expression type, a referential expression type, an omitted expression type, a grammatical error type, and a pause expression type; The natural expression type refers to the expression type in which the text is expressed using natural language expression; The referential expression type refers to the expression type in which a pronoun is used to refer to an entity word in a text; The grammatical error type refers to the type of expression in which there are grammatical errors and / or expression errors in the text; The omitted expression type refers to an expression type in which key entity words are omitted in the text; The pause expression type refers to an expression type in which there is an expression association between the preceding and following texts.

18. The device according to claim 17, wherein: Also includes: A first collecting unit is used to collect first training data according to the imperative expression type, wherein the first training data includes a training text expressed according to a command expression structure corresponding to the imperative expression type and a type label corresponding to the imperative expression type; A second collecting unit is used to collect second training data according to the non-imperative expression type, wherein the second training data includes a training text collected according to the non-imperative expression type and a type label corresponding to the non-imperative expression type; A model training unit is used to train and obtain the text classifier based on the first training data and the second training data.

19. The device according to any one of claims 11 to 18, wherein: The type determination unit comprises: A complexity determination module, used to determine the text complexity corresponding to the text information; A first determination module, configured to determine that the imperative expression type is the text expression type corresponding to the voice information if the text complexity is less than or equal to a complexity threshold; The second determination module is used to determine that the non-imperative expression type is the text expression type corresponding to the voice information if the text complexity is greater than the complexity threshold.

20. The device according to any one of claims 11 to 19, wherein: The understanding output unit comprises: A program determination module, used to determine a target application corresponding to the understanding result information from at least one application supported by the vehicle-mounted terminal; A result generation module, used for generating an interaction result corresponding to the understanding result information according to the target application; A result output module is used to play the interaction result through the target application of the vehicle terminal.

21. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 10.

22. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-10.

23. A computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 10.

Images