Voice interaction method, device, equipment, medium and product

Abstract

The invention discloses a voice interaction method and device, equipment, a medium and a product, and relates to the technical field of artificial intelligence, and the method comprises the steps: responding to the received user voice content sent by a vehicle machine, and recognizing a domain classification corresponding to the user voice content; calling a target language model according to the domain classification, and generating or determining reply content corresponding to the user voice content through the target language model; the reply content is sent to the vehicle machine, so that the vehicle machine provides the reply content to the user, the domain classification corresponding to the user voice content is recognized, and then the target language model is called according to the domain classification, so that the target language model is adaptively called according to the user voice content, and the user experience is improved. The advantages of different language models can be fully played, and the voice interaction experience of the user is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a voice interaction method, device, equipment, medium and product. Background Art

[0002] With the continuous development of large model technology, the capabilities of large models are now very powerful and can answer a variety of questions. They have been widely used in various industries.

[0003] In-car voice currently mainly uses traditional models to understand what users say and interact with users by voice. Traditional models perform well in traditional fields such as vehicle control, navigation, and multimedia, but they are not effective in fields with more diverse answers such as itinerary planning, scenic spot recommendations, and film and television recommendations. On the contrary, large models can give rich results in these fields with more divergent answers, but large models do not perform as well as traditional models in traditional fields. At present, there is no clear solution for what model to use to handle in-car voice interaction, which affects the user's voice interaction experience.

[0004] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0005] The main purpose of this application is to provide a voice interaction method, device, equipment, medium and product, aiming to improve the user's voice interaction experience.

[0006] To achieve the above object, the present application provides a voice interaction method, which is applied to a voice cloud, comprising:

[0007] In response to receiving the user voice content sent by the vehicle computer, identifying the field classification corresponding to the user voice content;

[0008] Calling a target language model according to the domain classification, and generating or determining a reply content corresponding to the user voice content through the target language model;

[0009] The reply content is sent to the vehicle computer so that the vehicle computer provides the reply content to the user.

[0010] In one embodiment, the target language model includes a first language model and / or a second language model, and the step of calling the target language model according to the domain classification includes:

[0011] Identify whether the field classification is a target field;

[0012] If the domain is classified as a target domain, calling the first language model;

[0013] If the domain is classified as a non-target domain, the second language model is called.

[0014] In one embodiment, the step of generating or determining the reply content corresponding to the user voice content by using the target language model includes:

[0015] Reading the conversation history context information through the target language model, and generating or determining reply data according to the conversation history context information and the user voice content;

[0016] Selecting or determining a message template corresponding to the reply data from a pre-configured data template;

[0017] The reply data is fused and encapsulated according to the message template to obtain the reply content.

[0018] In one embodiment, after the step of generating or determining the reply content corresponding to the user voice content by using the target language model, the step further includes:

[0019] The conversation history context information is updated according to the user voice content and / or reply content.

[0020] In addition, to achieve the above purpose, the present application also proposes a voice interaction method, which is applied to a vehicle computer and includes:

[0021] Get user voice content;

[0022] Sending the user voice content to a voice cloud, so that the voice cloud, in response to receiving the user voice content, identifies the domain classification corresponding to the user voice content, calls a target language model according to the domain classification, and generates or determines a reply content corresponding to the user voice content through the target language model;

[0023] Receive the reply content sent by the voice cloud, and provide the reply content to the user.

[0024] In one embodiment, the step of acquiring the user voice content includes: collecting the original user voice through a sensor, performing text recognition on the original user voice, and obtaining the user voice content.

[0025] In one embodiment, the step of providing the reply content to the user includes: presenting the reply content through interface display and / or voice output.

[0026] In addition, to achieve the above purpose, the present application also proposes a voice interaction device, which is applied to a voice cloud, including:

[0027] A recognition module, configured to, in response to receiving user voice content sent by the vehicle computer, identify a field classification corresponding to the user voice content;

[0028] A calling module, used to call a target language model according to the domain classification, and generate or determine a reply content corresponding to the user voice content through the target language model;

[0029] The sending module is used to send the reply content to the vehicle computer so that the vehicle computer provides the reply content to the user.

[0030] In addition, to achieve the above-mentioned purpose, the present application also proposes a voice interaction device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the voice interaction method described above.

[0031] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the voice interaction method described above are implemented.

[0032] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the voice interaction method described above.

[0033] One or more technical solutions proposed in this application have at least the following technical effects:

[0034] In response to receiving user voice content sent by the vehicle computer, identifying the domain classification corresponding to the user voice content; calling the target language model according to the domain classification, and generating or determining reply content corresponding to the user voice content through the target language model; sending the reply content to the vehicle computer so that the vehicle computer provides the reply content to the user. By identifying the domain classification corresponding to the user voice content and then calling the target language model according to the domain classification, the target language model is adaptively called according to the user voice content, which can give full play to the advantages of different language models and enhance the user's voice interaction experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0036] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0037] Figure 1 A flowchart of the first embodiment of the voice interaction method of the present application is provided;

[0038] Figure 2 A flowchart of the second embodiment of the voice interaction method of the present application is provided;

[0039] Figure 3 This is a schematic diagram of the overall process in the embodiment of the present application;

[0040] Figure 4 A flowchart of the third embodiment of the voice interaction method of the present application is provided;

[0041] Figure 5 This is a schematic diagram of the module structure of the voice interaction device according to an embodiment of the present application;

[0042] Figure 6 This is a schematic diagram of the device structure of the hardware operating environment involved in the voice interaction method in the embodiment of the present application.

[0043] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0044] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0045] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0046] The main solution of the embodiment of the present application is: in response to receiving user voice content sent by the vehicle computer, identifying the domain classification corresponding to the user voice content; calling the target language model according to the domain classification, and generating or determining the reply content corresponding to the user voice content through the target language model; sending the reply content to the vehicle computer so that the vehicle computer provides the reply content to the user, by identifying the domain classification corresponding to the user voice content, and then calling the target language model according to the domain classification, it is achieved that the target language model is adaptively called according to the user voice content, which can give full play to the advantages of different language models and enhance the user's voice interaction experience.

[0047] In this embodiment, for the convenience of description, the following description is made with the voice interaction device as the execution subject.

[0048] With the continuous development of large model technology, the capabilities of large models are now very powerful and can answer a variety of questions. They have been widely used in various industries.

[0049] At present, in-vehicle voice mainly uses traditional models to understand what users say and interact with users by voice. Traditional models have excellent performance in traditional fields such as vehicle control, navigation, and multimedia, but they are not effective in fields with more diverse answers such as itinerary planning, scenic spot recommendations, and film and television recommendations. On the contrary, large models can give rich results in these fields where the answers are more divergent. In the embodiments of the present application, by organically integrating traditional models with large models, users can be provided with rich and colorful answers.

[0050] This application provides a solution that dynamically decides whether to use the results of the traditional voice model or the understanding results of the large model according to the field classification corresponding to the text content of the user's voice recognition, organically integrating the advantages of the traditional model and the large model, and learning from each other's strengths to achieve the best response to the user's conversation. This application also stores the historical conversations during the user's voice interaction and passes them to the large model, so that the large model can better understand and give the most relevant response to the question in combination with the historical context, which can significantly improve the user's voice interaction experience.

[0051] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a voice interaction device, etc. The following takes the voice interaction device as an example to illustrate this embodiment and the following embodiments.

[0052] Based on this, the present application embodiment provides a voice interaction method, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the voice interaction method of the present application.

[0053] In this embodiment, the voice interaction method includes steps S10 to S30:

[0054] Step S10, in response to receiving the user voice content sent by the vehicle computer, identifying the field classification corresponding to the user voice content;

[0055] Specifically, in-vehicle voice interaction can enhance the driving experience by providing an intuitive and safe way of interaction. In an embodiment of the present application, the user's voice can be collected through the built-in sensors of the vehicle (such as a microphone, etc.), and the user's voice can be converted into text through the voice recognition function to obtain the user's voice content.

[0056] Optionally, in the embodiments of the present application, the vehicle-mounted computer sends the obtained user voice content to the voice cloud. After receiving the user voice content, the voice cloud can input the user voice content into the domain classification module, and the domain classification module can output the specific domain classification corresponding to the user voice content.

[0057] Optionally, in the domain classification module, the user voice content can be classified by a pre-trained domain classification model. The training process of the domain classification model includes: obtaining a text data set containing known domain classifications, performing data preprocessing on the obtained text data set, such as cleaning, tokenization, normalization, and stop word removal. Through cleaning, irrelevant characters, HTML tags, special symbols, etc. can be removed; the text is tokenized using nltk or other tools; the normalization process includes converting to lowercase, stemming, or lemmatization; common stop words such as "de" (of), "shi" (is), etc. are removed to obtain the preprocessed text data. Then, feature extraction is performed on the preprocessed text data. In this process, a suitable feature representation method is selected, such as TF-IDF (Term Frequency-Inverse Document Frequency), Word Embeddings, etc. Then, through vectorization processing, the text is converted into a numerical vector to facilitate processing by machine learning models. In terms of model selection, Naive Bayes, SVM (Support Vector Machine), Random Forest, and deep learning models can be selected, such as CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), and BERT (Bidirectional Encoder Representations from Transformers). The data set is divided into a training set, a validation set, and a test set. The selected model is trained using the training set, and the model parameters are tuned on the validation set to avoid overfitting. The performance of the model is evaluated on an independent test set. Common evaluation metrics include accuracy, precision, recall, etc. By analyzing the cases of domain classification errors, the weaknesses of the model are understood, and the model architecture or feature engineering is adjusted according to the performance of the model on the test set. Finally, an optimized domain classification model is obtained.

[0058] Step S20: Invoke the target language model according to the domain classification, and generate or determine a reply content corresponding to the user voice content through the target language model;

[0059] Furthermore, after identifying the domain classification corresponding to the user's voice content, the target language model can be called according to the domain classification, and the reply content corresponding to the user's voice content can be generated or determined through the target language model.

[0060] Optionally, if the domain classification is a target domain such as navigation or vehicle control, the voice cloud calls the traditional model (i.e., the first language model) and uses the reply content of the traditional model to send it to the vehicle computer. At the same time, the current conversation information is saved in the conversation history context module for subsequent use.

[0061] Optionally, if the domain classification is a non-target domain such as itinerary planning, variety show search, etc., the voice cloud calls the big model (i.e., the second language model). When calling the big model, the previous user conversation history (obtained from the historical context module) information is passed to the big model, so that the big model can combine the historical conversation with the current voice recognition information for in-depth understanding and give a strongly relevant answer.

[0062] Step S30, sending the reply content to the vehicle computer, so that the vehicle computer provides the reply content to the user.

[0063] Furthermore, after calling the target language model according to the domain classification and generating or determining the reply content corresponding to the user's voice content through the target language model, the reply content can be sent to the vehicle computer so that the vehicle computer can provide the reply content to the user.

[0064] Optionally, after the voice cloud generates or determines the reply content by calling the target language model, the reply content is returned to the vehicle computer, and the reply content is provided to the user through the interface display and / or voice broadcast in the vehicle computer, thereby realizing in-vehicle voice interaction.

[0065] For example, suppose the user says: "What are some attractions worth visiting nearby?" The car system captures the user's voice through the microphone, performs noise reduction processing, and converts the voice into text: "What are some attractions worth visiting nearby?" The domain classification module of the voice cloud determines that the domain classification corresponding to the user's voice content is "travel-attraction recommendation", and determines that the domain classification belongs to a non-target domain. Then, the big model can be called for processing to query nearby tourist attractions information and related reviews. The reply content is generated or determined by the big model and sent to the car. The speaker in the car plays the synthesized voice: "There are several good attractions nearby for you to choose from. You can consider visiting [Attraction A], which has beautiful scenery; or go to [Attraction B] to experience local special activities; if you are interested in history and culture, [Attraction C] is also a good choice." The on-board display screen shows specific information about the relevant attractions.

[0066] This embodiment adopts the above scheme, specifically by responding to user voice content sent by the vehicle computer, identifying the domain classification corresponding to the user voice content; calling the target language model according to the domain classification, and generating or determining the reply content corresponding to the user voice content through the target language model; sending the reply content to the vehicle computer, so that the vehicle computer provides the reply content to the user, by identifying the domain classification corresponding to the user voice content, and then calling the target language model according to the domain classification, it is realized that the target language model is adaptively called according to the user voice content, which can give full play to the advantages of different language models and enhance the user's voice interaction experience.

[0067] Based on the first embodiment of the present application, a second embodiment of the present application is proposed. In the second embodiment of the present application, the same or similar contents as those of the above-mentioned embodiment 1 can be referred to the above introduction and will not be repeated later.

[0068] On this basis, please refer to Figure 2 , step S20 includes steps S201 to S202:

[0069] S201: calling a target language model according to the domain classification;

[0070] Optionally, the target language model includes a first language model and / or a second language model.

[0071] Optionally, the step of calling the target language model according to the domain classification includes:

[0072] Identify whether the field classification is a target field;

[0073] If the domain is classified as a target domain, calling the first language model;

[0074] If the domain is classified as a non-target domain, the second language model is called.

[0075] Optionally, the field classification in the embodiment of the present application may include a primary classification and a secondary classification, for example: the primary classification includes travel, entertainment, information and other fields, among which travel includes secondary classifications such as location area, country recommendation, foreign location, transportation method and attraction recommendation, and entertainment includes secondary classifications such as TV series introduction, TV series search, movie introduction, movie search, song introduction, song search, singer introduction, album introduction, album search, variety show introduction and variety show search.

[0076] Reference Figure 3 , Figure 3 It is a schematic diagram of the overall process in the embodiment of the present application, such as Figure 3As shown, after determining the domain classification of the user's voice content through the domain classification module, it is possible to determine whether the output domain classification belongs to the target domain, and then determine the language model that needs to be called.

[0077] Optionally, due to the excellent performance of traditional models in traditional fields such as vehicle control, navigation, and multimedia, it is not effective in fields with diverse answers such as itinerary planning, scenic spot recommendation, and film and television recommendation. On the contrary, the large model can give rich results in fields where these answers are more divergent. Therefore, in an embodiment of the present application, if the field classification of the user's voice content is determined to be the target field, the first language model (i.e., the traditional model) is called, and if the field classification of the user's voice content is determined to be the non-target field, the second language model (i.e., the large model) is called; in other embodiments, if the user's voice content covers the content of the target field and the non-target field, the second language model can be preferentially applied, and in the case of sufficient computing resources, the user's voice content can also be decomposed, and the content involving the target field is called by the first language model for processing, and the content involving the non-target field is called by the second language model for processing, and the reply data of different language models are further fused and processed to obtain the final reply content feedback to the user, so that the advantages of the traditional model and the large model can be fully utilized, and the model fusion and complementary advantages can be achieved, thereby improving the user's voice interaction experience.

[0078] Optionally, the voice cloud service will be connected to the resource center service. All the "domain classifications" corresponding to a query (i.e., user voice content) are configured in the resource center service. In the resource center service, it is also configured whether a certain domain classification calls the big model.

[0079] Optionally, according to the iteration of the big model capabilities, the configuration in the resource center can be continuously adjusted. For example, a certain field classification previously used traditional model results, and then it was adjusted to use the big model results.

[0080] Optionally, whether to call a traditional model or a large model for a certain domain classification output by the domain classification module is dynamically configurable in the voice cloud and can be dynamically adjusted according to the actual response effect. For example, after subsequent technical iterations of the model, for the field of "trip planning", if the response effect of the traditional model is better than that of the large model, the domain classification corresponding to "trip planning" can be set in real time in the voice cloud to use the traditional model results instead of the large model results, so as to achieve the ultimate user voice interaction optimal experience.

[0081] S202: Generate or determine reply content corresponding to the user voice content through the target language model.

[0082] Optionally, the step of generating or determining the reply content corresponding to the user voice content by using the target language model includes:

[0083] Reading the conversation history context information through the target language model, and generating or determining reply data according to the conversation history context information and the user voice content;

[0084] Selecting or determining a message template corresponding to the reply data from a pre-configured data template;

[0085] The reply data is fused and encapsulated according to the message template to obtain the reply content.

[0086] Optionally, in an embodiment of the present application, the user conversation history is stored through a conversation history context module. When the big model and / or traditional model is called, the conversation history context information stored in the history context module can be sent to the big model and / or traditional model, so that the big model and / or traditional model can combine the conversation history context information to deeply understand the user's voice content, and then generate or determine more accurate reply data.

[0087] Optionally, as another implementation method, when computing power is sufficient, the big model and the traditional model can be called at the same time, and then the big model data generated by the big model can be compared with the traditional model data generated by the traditional model, and the data with better response effect can be selected as the response data. The big model data and the traditional model data can also be integrated to obtain integrated response data, so as to further achieve the optimal user voice interaction experience.

[0088] Optionally, a data fusion and encapsulation module is also provided in the embodiment of the present application, and several message templates are pre-set in the voice cloud. Since the messages of user voice content in different fields are usually different, the message structure of "trip planning" is in format A, the message of "attraction recommendation" is in format B, and the message of "song recommendation" is in format C. If the field classification of the user voice content is "trip planning", the key fields in the result of the large model are assigned to the A template, and so on, which can facilitate subsequent business expansion. Optionally, the pre-configured message template corresponds to the field classification, and there can be multiple field classifications corresponding to one message template. According to the message template, the reply data output by the large model and / or the traditional model is fused and encapsulated to obtain the corresponding reply content, so that the reply data can be better presented on the car interface, and / or provided to the user through voice broadcast, thereby completing this round of voice interaction.

[0089] Optionally, after the step of generating or determining the reply content corresponding to the user voice content by using the target language model, the step further includes:

[0090] The conversation history context information is updated according to the user voice content and / or reply content.

[0091] Optionally, the user voice content and / or reply content of the current round of voice interaction is stored through the conversation history context module, so as to facilitate in-depth understanding in the next round of voice interaction in combination with the current round of conversation information, and then generate reply content for the next round of voice interaction that is strongly related to the current round of voice interaction, thereby improving the user's overall voice interaction experience.

[0092] Optionally, in the embodiment of the present application, it is also possible to collect feedback results from users on the reply content provided by the car computer, such as "satisfied" or "unsatisfied", or provide a specific scoring mechanism to collect user evaluations, and then optimize the target language model and / or message template configured for the corresponding field classification, and further update the large model and / or traditional model to improve the subsequent voice interaction experience.

[0093] This embodiment adopts the above scheme, specifically by identifying whether the domain classification is a target domain; if the domain classification is a target domain, calling the first language model; if the domain classification is a non-target domain, calling the second language model; reading the session history context information through the target language model, and generating or determining the reply data according to the session history context information and the user voice content; selecting or determining the message template corresponding to the reply data from the pre-configured data template; fusing and packaging the reply data according to the message template to obtain the reply content, and generating or determining the reply data according to the session history context information and the user voice content, so that the model combines the historical dialogue to generate the reply data strongly related to the user voice content; by adaptively calling the corresponding model, the advantages of the traditional model and the large model can be fully utilized to achieve model fusion and complementary advantages, thereby improving the user's voice interaction experience.

[0094] Based on any of the foregoing embodiments of the present application, a third embodiment of the present application is proposed. In the third embodiment of the present application, the same or similar contents as any of the foregoing embodiments can be referred to the above introduction and will not be described in detail later. Figure 4 The voice interaction method is applied to a vehicle computer, and includes steps A10 to A30:

[0095] A10: Obtain user voice content;

[0096] Optionally, the step of acquiring the user voice content includes:

[0097] The original user voice is collected by a sensor, and text recognition is performed on the original user voice to obtain the user voice content.

[0098] Optionally, in-vehicle voice interaction can enhance the driving experience by providing an intuitive and safe interaction method. In the embodiments of this application, the vehicle-mounted computer can collect the user's voice through built-in sensors (such as microphones), and perform noise reduction and echo cancellation on the collected audio signal to improve the accuracy of speech recognition. Through the speech recognition function, the user's voice can be converted into text to obtain the user's voice content.

[0099] Optionally, in the embodiments of this application, the vehicle-mounted computer sends the obtained user voice content to the voice cloud. After receiving the user voice content, the voice cloud can input the user voice content into the domain classification module, and the domain classification module can output the specific domain classification corresponding to the user voice content.

[0100] A20: Send the user voice content to the voice cloud, so that the voice cloud, in response to receiving the user voice content, recognizes the domain classification corresponding to the user voice content, calls the target language model according to the domain classification, and generates or determines the reply content corresponding to the user voice content through the target language model;

[0101] Optionally, in the domain classification module, the user voice content can be classified by a pre-trained domain classification model. Among them, the training process of the domain classification model includes: obtaining a text data set containing known domain classifications, performing data preprocessing on the obtained text data set, such as cleaning, tokenization, normalization, and stop word removal, etc. Through cleaning, irrelevant characters, HTML tags, special symbols, etc. can be removed; tokenize the text using nltk or other tools; the normalization process includes converting to lowercase, stemming or lemmatization, etc.; remove common stop words, such as "de", "shi", etc., to obtain the preprocessed text data. Then, feature extraction is performed on the preprocessed text data. In this process, a suitable feature representation method is selected, such as TF-IDF, Word Embeddings, etc. Then, through vectorization processing, the text is converted into a numerical vector to facilitate the processing of machine learning models. In terms of model selection, naive Bayes, support vector machine (SVM), random forest, and deep learning models (such as CNN, RNN, BERT) can be selected, etc. The data set is divided into a training set, a validation set, and a test set. The selected model is trained using the training set, and the model parameters are tuned on the validation set to avoid overfitting. Evaluate the model performance on an independent test set. Common evaluation metrics include accuracy, precision, recall, etc. By analyzing the cases of domain classification errors, understand the weaknesses of the model, and adjust the model architecture or feature engineering according to the performance of the model on the test set to finally obtain an optimized domain classification model.

[0102] Optionally, if the domain classification is a target domain such as navigation or vehicle control, the voice cloud calls the traditional model (i.e., the first language model) and uses the reply content of the traditional model to send it to the vehicle computer. At the same time, the current conversation information is saved in the conversation history context module for subsequent use.

[0103] Optionally, if the domain classification is a non-target domain such as itinerary planning, variety show search, etc., the voice cloud calls the big model (i.e., the second language model). When calling the big model, the previous user conversation history (obtained from the historical context module) information is passed to the big model, so that the big model can combine the historical conversation with the current voice recognition information for in-depth understanding and give a strongly relevant answer.

[0104] A30: Receive the reply content sent by the voice cloud, and provide the reply content to the user.

[0105] Optionally, after the voice cloud generates or determines the reply content by calling the target language model, the reply content is returned to the vehicle computer. The vehicle computer receives the reply content sent by the voice cloud and provides the reply content to the user.

[0106] Optionally, the step of providing the reply content to the user includes:

[0107] The reply content is presented through interface display and / or voice output.

[0108] Optionally, in the embodiment of the present application, text-to-speech (TTS) technology can be used to convert the generated text into speech, and the sound quality of the synthesized speech can be further adjusted to make it sound more natural and smooth. The synthesized speech is played through the vehicle speaker to convey the reply content to the user. In some cases, relevant text information or images can also be displayed on the vehicle display to enhance the user experience.

[0109] Optionally, during the speech synthesis process, rule-based methods, statistical methods and deep learning methods (such as WaveNet and Tacotron, etc.) can be used to adjust the emotion and intonation of the speech according to the context, so that the synthesized speech is more natural and expressive; during the speech output process, the user's pre-set timbre can be used for speech synthesis, or a suitable timbre can be automatically selected from a speech library for speech broadcast, so that the broadcast voice is more in line with user habits, thereby further improving the user's voice interaction experience.

[0110] Optionally, during the interface display process, a short title or summary can be displayed based on the reply content to summarize the main information, or more detailed text content can be displayed, such as address, description, steps, etc. For travel-related field classifications, maps and / or route maps can be displayed to help users better understand geographic location information, and related pictures, such as photos of scenic spots, can also be displayed. By providing interactive elements such as buttons, users are allowed to perform further operations, such as viewing details or starting navigation, etc., and evaluation-type interactive elements can also be used to quickly provide feedback on the results of this round of voice interaction, which is used to optimize the selection and processing methods of large models and / or traditional models.

[0111] For example, suppose the user says, "What are some nearby attractions worth visiting?" The car system captures the user's voice through a microphone, performs noise reduction processing, and converts the voice into text: "What are some nearby attractions worth visiting?" The domain classification module of the voice cloud determines that the domain classification corresponding to the user's voice content is "travel-attraction recommendation", and determines that the domain classification belongs to a non-target domain. Then, the big model can be called for processing to query nearby tourist attractions information and related reviews. The reply content is generated or determined by the big model and sent to the car. The speaker in the car plays the synthesized voice: "There are several good attractions nearby for you to choose from. You can consider visiting [Attraction A], which has beautiful scenery; or go to [Attraction B] to experience local special activities; if you are interested in history and culture, [Attraction C] is also a good choice." The in-vehicle display screen displays specific information about the relevant attractions.

[0112] This embodiment adopts the above scheme, specifically by acquiring user voice content; sending the user voice content to the voice cloud, so that the voice cloud responds to receiving the user voice content, identifies the domain classification corresponding to the user voice content, calls the target language model according to the domain classification, and generates or determines the reply content corresponding to the user voice content through the target language model; receives the reply content sent by the voice cloud, and provides the reply content to the user, by identifying the domain classification corresponding to the user voice content, and then calling the target language model according to the domain classification, it is realized that the target language model is adaptively called according to the user voice content, which can give full play to the advantages of different language models and enhance the user's voice interaction experience.

[0113] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the voice interaction method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0114] This application also provides a voice interaction device, please refer to Figure 5 , the voice interaction device is applied to the voice cloud, including:

[0115] A recognition module, configured to, in response to receiving user voice content sent by the vehicle computer, identify a field classification corresponding to the user voice content;

[0116] A calling module, used to call a target language model according to the domain classification, and generate or determine a reply content corresponding to the user voice content through the target language model;

[0117] The sending module is used to send the reply content to the vehicle computer so that the vehicle computer provides the reply content to the user.

[0118] The voice interaction device provided by the present application adopts the voice interaction method in the above embodiment, which can solve the technical problems of voice interaction. Compared with the prior art, the beneficial effects of the voice interaction device provided by the present application are the same as the beneficial effects of the voice interaction method provided by the above embodiment, and other technical features in the voice interaction device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0119] The present application provides a voice interaction device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the voice interaction method in the above-mentioned embodiment one.

[0120] Reference below Figure 6 , which shows a schematic diagram of the structure of a voice interaction device suitable for implementing the embodiments of the present application. The voice interaction device in the embodiments of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 6 The voice interaction device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0121] like Figure 6As shown, the voice interaction device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the voice interaction device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the voice interaction device to communicate with other devices wirelessly or by wire to exchange data. Although the figure shows a voice interaction device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or provided instead.

[0122] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0123] The voice interaction device provided by the present application adopts the voice interaction method in the above embodiment to solve the technical problem of voice interaction. Compared with the prior art, the beneficial effects of the voice interaction device provided by the present application are the same as the beneficial effects of the voice interaction method provided by the above embodiment, and the other technical features in the voice interaction device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0124] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0125] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0126] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, wherein the computer-readable program instructions are used to execute the voice interaction method in the above-mentioned embodiment.

[0127] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, 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 above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0128] The computer-readable storage medium may be included in the voice interaction device; or may exist independently without being assembled into the voice interaction device.

[0129] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the voice interaction device, the voice interaction device: in response to receiving the user voice content sent by the vehicle computer, identifies the domain classification corresponding to the user voice content; calls the target language model according to the domain classification, and generates or determines the reply content corresponding to the user voice content through the target language model; sends the reply content to the vehicle computer so that the vehicle computer provides the reply content to the user. By identifying the domain classification corresponding to the user voice content and then calling the target language model according to the domain classification, the target language model is adaptively called according to the user voice content, which can give full play to the advantages of different language models and enhance the user's voice interaction experience.

[0130] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0131] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0132] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0133] The readable storage medium provided in this application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned voice interaction method, and can solve the technical problems of voice interaction. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the voice interaction method provided in the above-mentioned embodiment, and will not be repeated here.

[0134] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned voice interaction method when executed by a processor.

[0135] The computer program product provided by this application can solve the technical problem of voice interaction. Compared with the prior art, the beneficial effects of the computer program product provided by this application are the same as the beneficial effects of the voice interaction method provided by the above embodiment, which will not be repeated here.

[0136] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A voice interaction method, characterized in that: The method is applied to a voice cloud, and includes: In response to receiving the user voice content sent by the vehicle computer, identifying the field classification corresponding to the user voice content; Calling a target language model according to the domain classification, and generating or determining a reply content corresponding to the user voice content through the target language model; The reply content is sent to the vehicle computer so that the vehicle computer provides the reply content to the user.

2. The method according to claim 1, characterized in that The target language model includes a first language model and / or a second language model, and the step of calling the target language model according to the domain classification includes: Identify whether the field classification is a target field; If the domain is classified as a target domain, calling the first language model; If the domain is classified as a non-target domain, the second language model is called.

3. The method according to claim 1, characterized in that The step of generating or determining the reply content corresponding to the user voice content by using the target language model includes: Reading the conversation history context information through the target language model, and generating or determining reply data according to the conversation history context information and the user voice content; Selecting or determining a message template corresponding to the reply data from a pre-configured data template; The reply data is fused and encapsulated according to the message template to obtain the reply content.

4. The method according to claim 3, characterized in that After the step of generating or determining the reply content corresponding to the user voice content by using the target language model, the following step further includes: The conversation history context information is updated according to the user voice content and / or reply content.

5. A voice interaction method, characterized in that: The method is applied to a vehicle computer, and includes: Get user voice content; Sending the user voice content to a voice cloud, so that the voice cloud, in response to receiving the user voice content, identifies the domain classification corresponding to the user voice content, calls a target language model according to the domain classification, and generates or determines a reply content corresponding to the user voice content through the target language model; Receive the reply content sent by the voice cloud, and provide the reply content to the user.

6. The method according to claim 5, characterized in that The step of acquiring the user voice content includes: collecting the original user voice through a sensor, performing text recognition on the original user voice, and obtaining the user voice content; The step of providing the reply content to the user includes: presenting the reply content through interface display and / or voice output.

7. A voice interaction device, characterized in that: The device is applied to a voice cloud, and includes: A recognition module, configured to, in response to receiving user voice content sent by the vehicle computer, identify a field classification corresponding to the user voice content; A calling module, used to call a target language model according to the domain classification, and generate or determine a reply content corresponding to the user voice content through the target language model; The sending module is used to send the reply content to the vehicle computer so that the vehicle computer provides the reply content to the user.

8. A voice interaction device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the voice interaction method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the voice interaction method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the voice interaction method according to any one of claims 1 to 6 are implemented.

Images