Optimization method, voice interaction method, server and computer readable storage medium

Abstract

The invention discloses a model optimization method for vehicle voice interaction, a voice interaction method, a server and a computer readable storage medium. The method comprises the following steps: firstly, performing rejection transformation processing on an original training data set of a model to obtain a first training data set and a second training data set; thirdly, obtaining a third training data set according to the original training data set and the original log data set; and finally, training the model according to the original training data set, the first training data set, the second training data set and the third training data set so as to improve the rejection capability of the model. Therefore, by performing rejection transformation processing on the original training data set and performing specific processing according to the original training data set and the original log data set, the number of rejection training data sets is increased, the breadth of the rejection training data sets is widened, and the rejection capability of the model is improved through the rejection training data; thus, it is possible to prevent the model from erroneously executing requests that may cause dangers.

Description

Technical Field

[0001] The present invention relates to the field of voice interaction technology, and in particular to an optimization method for a model used for vehicle voice interaction, a voice interaction method, a server and a computer-readable storage medium. Background Art

[0002] In-vehicle intelligent systems often train their rejection capabilities by inputting rejection training data to improve the system's ability to reject inappropriate or invalid inputs in practical applications. However, due to the insufficient amount of rejection training data and the low breadth of rejection training data, the rejection capability of in-vehicle intelligent systems is poor. Summary of the invention

[0003] The present application provides a method for optimizing a model for vehicle voice interaction, a voice interaction method, a server, and a computer-readable storage medium.

[0004] The present application provides an optimization method for a model for vehicle voice interaction, the method comprising:

[0005] Performing rejection transformation processing on the original training data set of the model to obtain a first training data set and a second training data set;

[0006] Obtaining a third training data set according to the original training data set and the original log data set;

[0007] The model is trained according to the original training data set, the first training data set, the second training data set and the third training data set to improve the rejection capability of the model.

[0008] In this way, the server performs a rejection transformation process on the original training data set of the model to obtain a first training data set and a second training data set. Then, the server obtains a third training data set based on the original training data set and the original log data set. Finally, the server trains the model based on the original training data set, the first training data set, the second training data set, and the third training data set to improve the rejection capability of the model. In this way, by performing a rejection transformation process on the original training data set and performing specific processing on the original training data set and the original log data set, the number of rejection training data sets is increased and the breadth of the rejection training data sets is broadened, and the rejection capability of the model is improved through these rejection training data, thereby preventing the model from erroneously executing requests that may cause danger.

[0009] In some embodiments, the reject-recognition transformation process is performed on the original training data set of the model to obtain the first training data set and the second training data set, including:

[0010] Based on a preset test method, the first training data set is obtained according to a preset rejection description, an original rejection data subset and an original valid data subset in the original training data set.

[0011] In this way, based on the preset test method, the server obtains the first training data set according to the preset rejection description, the original rejection data subset and the original valid data subset in the original training data set. In this way, the server uses the preset test method and the preset rejection description, and combines the original rejection data subset and the original valid data subset in the original training data set to obtain the first training data set, which can help the model better understand under what circumstances it should refuse to process the input, and improve the security and reliability of the model in practical applications.

[0012] In some embodiments, the original valid data subset includes a plurality of original valid data, and the first training data set is obtained based on the preset test method according to the original training data set and the preset rejection description, including:

[0013] Based on a preset granularity, segment each of the original valid data in the original valid data subset to obtain a plurality of target rejected data, wherein each of the target rejected data corresponds to one of the original valid data;

[0014] According to the preset rejection description and the plurality of target rejection data, a plurality of temporary rejection data sets are obtained;

[0015] The temporary rejected data set is tested according to the original rejected data subset to confirm that at least one of the temporary rejected data sets is added to the first training data set.

[0016] In this way, the server performs segmentation processing on each original valid data in the original valid data subset based on the preset granularity to obtain multiple target rejection data, wherein each target rejection data corresponds to one original valid data. Then, the server obtains multiple temporary rejection data sets based on the preset rejection description and the multiple target rejection data. Finally, the server performs test processing on the temporary rejection data set based on the original rejection data subset to confirm that at least one of the temporary rejection data sets is added to the first training data set. In this way, through meticulous data processing and testing, it is ensured that the generated first training data set can effectively train the model, so that the model can run more safely and reliably in practical applications.

[0017] In some embodiments, the target rejection data includes at least one sentence, and the multiple temporary rejection data sets are obtained according to the preset rejection description and the target rejection data, including:

[0018] The preset rejection description is placed at a segmentation position in each of the target rejection data to obtain a plurality of temporary rejection data sets, wherein the segmentation position is a connection position of each sentence in the target rejection data.

[0019] In this way, the server places the preset rejection description at the segmentation position of each target rejection data, and obtains multiple temporary rejection data sets, where the segmentation position is the connection position of each sentence in the target rejection data. In this way, the server places the preset rejection description at the segmentation position of each first rejection data, and creates multiple new data sets, namely temporary rejection data sets, so as to generate a first training data set according to the temporary rejection data set to train the model and improve the rejection ability of the model in practical applications.

[0020] In some embodiments, the reject-recognition transformation process is performed on the original training data set of the model to obtain the first training data set and the second training data set, including:

[0021] The original valid data subset is transformed to obtain the second training data set.

[0022] In this way, the server transforms the original valid data subset to obtain a second training data set. In this way, the second training data set is obtained, which can help the model better understand under what circumstances it should refuse to process input, thereby improving the security and reliability of the model in practical applications.

[0023] In some embodiments, the transforming the original valid data subset to obtain the second training data set includes:

[0024] Analyzing and processing the original valid data in the original valid data subset to obtain target information to be transformed;

[0025] Performing transformation processing on the target information to be transformed to obtain target error information;

[0026] The second training data set is obtained according to the target error information.

[0027] In this way, the server parses the original valid data in the original valid data subset to obtain the target information to be transformed. Then, the server transforms the target information to be transformed to obtain the target error information. Finally, the server obtains the second training data set based on the target error information. In this way, by replacing the target information to be transformed with the generated target error information to generate the second training data set, the model can be trained to better understand under what circumstances the input should be rejected, thereby improving the safety and reliability of the model in practical applications.

[0028] In some implementations, obtaining a third training data set according to the original training data set and the original log data set includes:

[0029] Based on a preset similarity algorithm, obtaining the similarity of each log data according to the log data in the original log data set and the original valid data subset in the original training data set;

[0030] The third training data set is determined according to a preset similarity threshold and the similarity.

[0031] In this way, the server obtains the similarity of each log data based on the log data in the original log data set and the original valid data subset in the original training data set based on the preset similarity algorithm. Then, the server determines the third training data set based on the preset similarity threshold and the similarity. In this way, log data that is dissimilar or abnormal to the original valid data subset is obtained, and these log data can be used to train the model to improve the model's ability to reject inappropriate or abnormal inputs in practical applications.

[0032] In some embodiments, the original valid data subset includes a plurality of original valid data, and the similarity of each log data is obtained based on the log data in the original log data set and the original valid data subset in the original training data set based on a preset similarity algorithm, including:

[0033] Determining a log data vector of the log data according to the preset similarity algorithm and the log data;

[0034] Determining an original valid data vector of the original valid data according to the preset similarity algorithm and the original valid data;

[0035] The similarity is obtained according to the log data vector and the original valid data vector.

[0036] In this way, the server determines the log data vector of the log data according to the preset similarity algorithm and the log data. Then, the server determines the original valid data vector of the original valid data according to the preset similarity algorithm and the original valid data. Finally, the server obtains the similarity according to the log data vector and the original valid data vector. In this way, the server can evaluate the similarity between the log data and the original valid data, so as to better understand the relationship between the data and provide a basis for subsequent data processing and analysis.

[0037] In some implementations, determining the log data vector of the log data according to the preset similarity algorithm and the log data includes:

[0038] Determine, based on the log data, a word vector in the log data;

[0039] Determine a target word according to the log data;

[0040] According to the target word and the original valid data subset, a word frequency-inverse document frequency weight of the target word is obtained;

[0041] A log data vector of the log data is obtained according to the word vector and the word frequency-inverse document frequency weight.

[0042] In this way, the server determines the word vector in the log data based on the log data. Subsequently, the server determines the target word based on the log data. Next, the server obtains the term frequency-inverse document frequency weight of the target word based on the target word and the original valid data subset. Finally, the server obtains the log data vector of the log data based on the term vector and the term frequency-inverse document frequency weight. In this way, a log data vector composed of the TF-IDF weight of the target word is obtained, which can effectively represent the characteristics and content of the log data, thereby better understanding the content and theme of the data for subsequent similarity calculations.

[0043] In some embodiments, the training of the model according to the original training data set, the first training data set, the second training data set, and the third training data set to improve the rejection capability of the model includes:

[0044] Obtaining a rejection training data set according to the second training data set and the third training data set;

[0045] Based on a preset ratio, the model is trained according to the original training data set, the first training data set and the rejection training data set to improve the rejection capability of the model.

[0046] In this way, the server obtains a rejection training data set based on the second training data set and the third training data set. Then, the server trains the model based on the original training data set, the first training data set and the rejection training data set based on a preset ratio to improve the rejection capability of the model. In this way, the server determines the usage of data in the original training data set, the first training data set and the rejection training data set based on a preset ratio to achieve a good training effect, thereby improving the model's ability to reject inappropriate or abnormal inputs in practical applications.

[0047] The embodiment of the present application provides a voice interaction method, which is based on a model optimized by the above-mentioned optimization method, and the method includes:

[0048] Receive voice requests forwarded by the vehicle;

[0049] The voice request is rejected according to the model to complete the voice interaction.

[0050] In this way, the server receives the voice request forwarded by the vehicle. Then, the server rejects the voice request according to the model to complete the voice interaction. In this way, based on the model trained by the above optimization method, the server can ensure that the voice interaction between the vehicle and the user is both safe and reliable.

[0051] An embodiment of the present application provides a server, which includes a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the above-mentioned optimization method or voice interaction method is implemented.

[0052] An embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the above-mentioned optimization method or voice interaction method are implemented.

[0053] Additional aspects and advantages of the embodiments of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0055] Figure 1 It is one of the flow diagrams of the optimization method of the model of certain embodiments of the present application;

[0056] Figure 2 This is the second flow chart of the optimization method of the model in certain embodiments of the present application;

[0057] Figure 3 This is the third flow chart of the optimization method of the model in certain embodiments of the present application;

[0058] Figure 4 This is a fourth flow chart of a method for optimizing a model according to certain embodiments of the present application;

[0059] Figure 5 This is a fifth flow chart of a method for optimizing a model according to certain embodiments of the present application;

[0060] Figure 6 This is the sixth flow chart of the optimization method of the model in certain embodiments of the present application;

[0061] Figure 7 This is the seventh flow chart of the optimization method of the model in certain embodiments of the present application;

[0062] Figure 8 This is the eighth flow chart of the optimization method of the model in certain embodiments of the present application;

[0063] Fig. 9 This is a ninth flow chart of a method for optimizing a model according to certain embodiments of the present application;

[0064] Fig.10 FIG10 is a flow chart of a method for optimizing a model according to certain embodiments of the present application;

[0065] Fig.11 It is a schematic diagram of the model training data processing flow of the voice interaction method in certain embodiments of the present application;

[0066] Fig.12 It is a flowchart of the voice interaction method of certain embodiments of the present application. DETAILED DESCRIPTION

[0067] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions from beginning to end. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the embodiments of the present application, and cannot be understood as limiting the embodiments of the present application.

[0068] In-vehicle intelligent systems usually train their rejection capabilities by inputting rejection training data to improve the system's ability to reject inappropriate or invalid inputs in actual applications. However, due to the insufficient amount of rejection training data and the low breadth of rejection training data, the rejection capability of the in-vehicle intelligent system is poor, which may reduce the safety of the in-vehicle intelligent system. In this way, the in-vehicle intelligent system may encounter the following problems in actual applications:

[0069] First, due to insufficient rejection training data, the in-vehicle intelligent system may not be able to accurately identify inappropriate or invalid inputs, resulting in misrecognition. For example, the in-vehicle intelligent system may mistakenly recognize the driver's request as a valid request and perform an incorrect operation.

[0070] Second, the low breadth of rejection training data may result in the inability of the in-vehicle intelligent system to cover all possible inappropriate or invalid input situations. As a result, the system may miss some inappropriate or invalid inputs, resulting in the inability to reject and process these inputs.

[0071] Based on the above questions, please refer to Figure 1 The present application embodiment provides a method for optimizing a model for vehicle voice interaction, the method comprising:

[0072] 01: Perform rejection transformation on the original training data set of the model to obtain the first training data set and the second training data set;

[0073] 02: Obtain a third training data set based on the original training data set and the original log data set;

[0074] 03: Train the model based on the original training data set, the first training data set, the second training data set, and the third training data set to improve the model's rejection capability.

[0075] The embodiment of the present application also provides a server, including a memory and a processor. The voice interaction method of the embodiment of the present application can be implemented by the server of the embodiment of the present application. Specifically, a computer program is stored in the memory, and the processor is used to perform a rejection transformation process on the original training data set of the model to obtain a first training data set and a second training data set. And a third training data set is obtained based on the original training data set and the original log data set. And the model is trained based on the original training data set, the first training data set, the second training data set, and the third training data set to improve the rejection capability of the model.

[0076] The embodiment of the present application also provides a model optimization device. The voice interaction method of the embodiment of the present application can be implemented by the model optimization device of the embodiment of the present application. Specifically, the model optimization device includes a rejection transformation module, a processing module and a training module. The rejection transformation module is used to perform rejection transformation processing on the original training data set of the model to obtain a first training data set and a second training data set. The processing module is used to obtain a third training data set based on the original training data set and the original log data set. The training module is used to train the model based on the original training data set, the first training data set, the second training data set and the third training data set to improve the rejection ability of the model.

[0077] Specifically, the original training dataset refers to a set of data used to train a machine learning model, which usually includes input samples and corresponding output labels, so that the model can learn how to extract features from the input data and make correct predictions or decisions. The original training dataset usually includes original valid data that can output correct responses and original rejected data that the system needs to reject, most of which are original valid data. In the context of in-vehicle intelligent systems, the original training dataset may include various driving scenarios, driver requests, vehicle status and other information, which are used to train the model to recognize and respond to different driving situations.

[0078] Rejection transformation refers to the process of processing the original training data set to generate a first training data set and a second training data set. Through this rejection transformation, the original valid data in the original training data set is converted into rejection data, which is used to train the machine learning model to improve its rejection ability. Rejection data usually includes specific requests and corresponding rejection labels, as well as various inappropriate or invalid inputs, and instructions on how the model should reject these inputs when encountering them. The rejection transformation method may include data enhancement, etc., to improve the model's ability to reject inappropriate or invalid inputs in practical applications.

[0079] The first training data set refers to a training data set obtained by performing rejection transformation processing on the preset rejection description and the original training data set.

[0080] The second training data set refers to the training data set obtained by directly transforming the original training data set, including input samples and corresponding output labels different from the first training data set, which is used to further improve the performance and generalization ability of the model. By using the second training data set, the model can learn more features and patterns, thereby improving its performance in practical applications.

[0081] The third training data set includes log data filtered from the original log data set. The log data has a low similarity with the data in the original training data set, and includes input samples and corresponding output labels that are different from the first two training data sets, which are used to further improve the performance and generalization ability of the model.

[0082] The model's recognition rejection capability refers to the model's ability to identify and refuse to process unfamiliar, abnormal or inappropriate inputs, thereby ensuring the safety and reliability of the model in practical applications.

[0083] The server performs a rejection transformation process on the original training data set of the model to obtain a first training data set and a second training data set.

[0084] Next, the server obtains a third training data set based on the original training data set and the original log data set. That is, the original log data set may include samples of user input in actual applications, and these samples may contain erroneous or abnormal inputs. By analyzing these log data, problems that the model may encounter in actual applications can be identified, and rejection samples can be generated accordingly.

[0085] Finally, the server trains the model based on the original training data set, the first training data set, the second training data set, and the third training data set to improve the model's rejection ability, so that the model learns to correctly reject inputs that are unfamiliar or inappropriate for the current task, rather than trying to make predictions or decisions that may lead to errors.

[0086] In summary, in the optimization method and server for the model of vehicle voice interaction provided in the embodiment of the present application, the server performs a rejection transformation process on the original training data set of the model to obtain a first training data set and a second training data set. Next, the server obtains a third training data set based on the original training data set and the original log data set. Finally, the server trains the model based on the original training data set, the first training data set, the second training data set, and the third training data set to improve the rejection capability of the model. In this way, by performing a rejection transformation process on the original training data set and performing specific processing on the original training data set and the original log data set, the number of rejection training data sets is increased and the breadth of the rejection training data sets is broadened, and the rejection capability of the model is improved through these rejection training data, thereby preventing the model from erroneously executing requests that may cause danger.

[0087] See also Figure 2 In some embodiments, step 01 (performing a rejection transformation process on the original training data set of the model to obtain a first training data set and a second training data set) includes:

[0088] 011: Based on a preset test method, a first training data set is obtained according to a preset rejection description, an original rejection data subset and an original valid data subset in an original training data set.

[0089] In some embodiments, the rejection transformation module is further configured to obtain a first training data set based on a preset test method, according to a preset rejection description, an original rejection data subset and an original valid data subset in the original training data set.

[0090] In some embodiments, the processor is further configured to obtain a first training data set based on a preset test method, according to a preset rejection description, an original rejection data subset and an original valid data subset in the original training data set.

[0091] Specifically, the predefined testing method refers to a pre-designed testing scheme that can be used to generate a first training data set, and the original rejected data subset and the original data in the original training data set are processed and converted to generate the first training data set.

[0092] A preset rejection description refers to a set of predefined rules or conditions used to identify and classify which input data should be rejected by the model. In the context of in-vehicle systems, these descriptions are usually based on an understanding of the task or problem, as well as prior knowledge of the dataset. Preset rejection descriptions ensure that the model can make correct rejection responses when faced with inappropriate or invalid input. For example, "If the request does not belong to the above functions, terminate the process and directly output a rejection label", "If the place name contained in the user's voice request does not exist in the system's place database, terminate the process and directly output a rejection label", and "If the user's voice request contains a request that the system does not support, terminate the process and directly output a rejection label" are all preset rejection descriptions.

[0093] The original rejection data subset refers to a portion of data separated from the original training data set, which is specifically used to test and evaluate the model's ability to reject recognition, including input data that the model should refuse to process, such as ambiguous requests, requests with incorrect formats, requests that exceed the scope of system functions, etc. The original rejection data subset can verify whether the model can correctly identify and refuse to process these inappropriate or invalid inputs, thereby verifying whether the model can correctly identify and refuse to process these inappropriate or invalid inputs. For example, in an in-vehicle system, the original rejection data subset may include non-existent place names, unsupported requests, ambiguous voice inputs, etc.

[0094] The original valid data subset refers to a part of the data separated from the original training data set, which can be correctly understood and processed, including input data that the model should process normally, such as clear requests, requests in the right format, requests within the scope of system functions, etc. The original valid data subset is used to train the model so that it can learn how to correctly respond to and process these valid inputs, so that the system can establish the ability to recognize and process valid inputs, so that it can effectively respond to user requests and requests in practical applications. For example, in an in-vehicle system, the original valid data subset may include navigating to a specific location, playing music, adjusting the temperature in the car, etc.

[0095] The server uses the preset test method and the preset rejection description to combine the original rejection data subset and the original valid data subset in the original training data set to generate the first rejection data set. For example, the original valid data is: "Please recognize and perform the corresponding operation according to the following driver's voice request: 'Please open the window.'", and after adding the preset rejection description to it, it may be: "Please recognize and perform the corresponding operation according to the following driver's voice request: 'Please open the window.' If the requested device is not a window, terminate the process and directly output the rejection label".

[0096] In this way, the server uses the preset test method and the preset rejection description, and combines the original rejection data subset and the original valid data subset in the original training data set to obtain the first training data set, which can help the model better understand under what circumstances it should refuse to process the input, and improve the security and reliability of the model in practical applications.

[0097] See also Figure 3 In some embodiments, the original valid data subset includes a plurality of original valid data, and step 011 (obtaining a first training data set based on a preset test method, according to the original training data set and a preset rejection description) includes:

[0098] 0111: Based on the preset granularity, each original valid data in the original valid data subset is segmented to obtain a plurality of target rejection data;

[0099] 0112: Obtain multiple temporary rejection data sets according to a preset rejection description and multiple target rejection data;

[0100] 0113: Perform test processing on the temporary rejection data set according to the original rejection data subset, and confirm to add at least one of the temporary rejection data sets to the first training data set.

[0101] In some embodiments, the rejection transformation module is further used to segment each original valid data in the original valid data subset based on a preset granularity to obtain a plurality of target rejection data. And to obtain a plurality of temporary rejection data sets based on the preset rejection description and the plurality of target rejection data. And to test the temporary rejection data sets based on the original rejection data subset to confirm that at least one of the temporary rejection data sets is added to the first training data set.

[0102] In some embodiments, the processor is further configured to segment each original valid data in the original valid data subset based on a preset granularity to obtain a plurality of target rejection data, and obtain a plurality of temporary rejection data sets based on the preset rejection description and the plurality of target rejection data, and perform a test process on the temporary rejection data set based on the original rejection data subset to confirm that at least one of the temporary rejection data sets is added to the first training data set.

[0103] Specifically, the preset granularity refers to the smallest unit of data segmentation or processing pre-defined by the server when processing the original valid data subset, which may be a paragraph, a sentence, a word, etc. In the implementation manner of the present application, the sentence is used as the preset granularity for description.

[0104] Target rejection data refers to the data obtained by segmenting the original valid data based on the preset granularity.

[0105] A temporary rejection data set refers to a set of temporary rejection data generated according to target rejection data and preset rejection descriptions. One target rejection data corresponds to one temporary rejection data set. In addition, the temporary rejection data in the temporary rejection data set may not necessarily be effective.

[0106] Based on the preset granularity, the server segments each original valid data in the original valid data subset to obtain multiple target rejection data, wherein each target rejection data corresponds to one original valid data. That is, the original valid data prompt is segmented at the sentence granularity (P_prompt = P_1+P_2+...+P_k, for example, P_1 is the first sentence of prompt), wherein P_1+P_2+...+P_k is one target rejection data. For example, the original valid data is: "Please recognize and perform the corresponding operation according to the following driver's voice request: 'Please open the window.'". Then the corresponding target rejection data generated may be: "[]Please recognize and perform the corresponding operation according to the following driver's voice request: []'Please open the window.' []".

[0107] Next, the server obtains multiple temporary rejection data sets based on the preset rejection description and multiple target rejection data. That is, the corresponding preset rejection description is added between the sentences in the target rejection data (for P_i, P_reject, P_prompt reject = P_1+P_2+...+P_i+P_reject+...+P_k). Continuing with the above example, the generated temporary rejection data set may include: "[If the request does not support the following functions, terminate the process and directly output the rejection label] Please identify and perform the corresponding operation according to the following driver's voice request: 'Please open the window.'". It should be noted that the corresponding preset rejection description can be added between every two sentences, at the beginning and at the end of the target rejection data, which will not be repeated here.

[0108] The temporary rejection dataset may also include: "Please recognize and perform the corresponding operation based on the following driver's voice request: 'Please open the window.' [If the request does not support the above functions, terminate the process and directly output the rejection label]".

[0109] Finally, the server tests the temporary rejection data set according to the original rejection data subset, and confirms to add at least one of the temporary rejection data sets to the first training data set. That is, each temporary rejection data set obtained above is tested using the data in the original rejection data subset to calculate the rejection accuracy. Then, according to specific needs, one or more temporary rejection data with the highest rejection accuracy are selected from each temporary rejection data set to add to the first training data set.

[0110] It should be noted that usually, a preset rejection description is added to the description task part at the beginning of the original valid data and to the few shot part of the original valid data example, which can significantly improve the rejection capability.

[0111] In this way, through careful data processing and testing, it is ensured that the generated first training data set can effectively train the model, so that the model can run more safely and reliably in practical applications.

[0112] See also Figure 4 In some embodiments, the target rejection data includes at least one sentence, and step 0112 (obtaining multiple temporary rejection data sets according to the preset rejection description and the target rejection data) includes:

[0113] 01121: Place the preset rejection description at the segmentation position in each target rejection data to obtain multiple temporary rejection data sets.

[0114] In some embodiments, the rejection transformation module is further used to place the preset rejection description at the segmentation position in each target rejection data, so as to obtain a plurality of temporary rejection data sets.

[0115] In some embodiments, the processor is further configured to place the preset rejection description at the segmentation position in each target rejection data, to obtain a plurality of temporary rejection data sets.

[0116] Specifically, the segmentation position refers to the position where the various sentences in the data are connected when processing the target rejection data. These positions are usually located after punctuation marks, spaces or other separators in the data, and are used to identify different parts or units in the data. For example, the original valid data is: "Please identify and perform the corresponding operations according to the following driver's voice request: 'Please open the window.'". Then the corresponding segmentation position is the position framed by the following "[]": "[]Please identify and perform the corresponding operations according to the following driver's voice request: []'Please open the window.' []".

[0117] The server places the preset rejection description at the segmentation position in each target rejection data to obtain multiple temporary rejection data sets, where the segmentation position is the connection position of each sentence in the target rejection data. That is, the preset rejection description is added at different segmentation positions of a target rejection data to obtain a temporary rejection data set, and multiple target rejection data obtain multiple temporary data sets.

[0118] In this way, the server places the preset rejection description at the segmentation position of each first rejection data, creates multiple new data sets, namely temporary rejection data sets, and generates a first training data set based on the temporary rejection data set to train the model and improve the rejection ability of the model in practical applications.

[0119] See also Figure 5 In some embodiments, step 01 (performing a rejection transformation process on the original training data set of the model to obtain a first training data set and a second training data set) includes:

[0120] 012: The original valid data subset is transformed to obtain a second training data set.

[0121] In some embodiments, the rejection transformation module is further used to transform the original valid data subset to obtain a second training data set.

[0122] In some embodiments, the processor is further configured to perform transformation processing on the original valid data subset to obtain a second training data set.

[0123] Specifically, transformation processing refers to adding noise, modifying data features, creating outliers, etc., with the aim of generating samples that the model should reject.

[0124] Adding noise refers to introducing random or artificial interference into speech data to simulate noisy environments in the real world. For example, background noise, such as wind noise, engine noise, or other people's conversations when the vehicle is driving, can be added to the driver's voice command to test the robustness of the model in different noise environments.

[0125] Modifying data features refers to intentionally changing some attributes or features of voice data to create new samples. For example, you can modify the capabilities of the in-car voice assistant, or change the pitch, speed or volume of the voice, or simulate different accents and pronunciations to generate new training samples.

[0126] Creating outliers refers to changing a portion of the original valid data in the original valid data subset to generate some samples that are significantly different from most of the speech data in the original training data set.

[0127] The server transforms the original valid data subset to obtain a second training data set.

[0128] In this way, a second training data set is obtained, which can help the model better understand under what circumstances it should refuse to process input, thereby improving the security and reliability of the model in practical applications.

[0129] See also Figure 6 In some embodiments, step 012 (transforming the original valid data subset to obtain a second training data set) includes:

[0130] 0121: Analyze and process the original valid data in the original valid data subset to obtain target information to be transformed;

[0131] 0122: Perform transformation processing on the target information to be transformed and obtain target error information;

[0132] 0123: Obtain a second training data set based on the target error information.

[0133] In some embodiments, the rejection modification processing module is used to parse the original valid data in the original valid data subset to obtain target information to be modified, and to modify the target information to be modified to obtain target error information, and to obtain a second training data set based on the target error information.

[0134] In some embodiments, the processor is further configured to parse the original valid data in the original valid data subset to obtain target information to be transformed, transform the target information to be transformed to obtain target error information, and obtain a second training data set based on the target error information.

[0135] Specifically, the target information to be modified refers to data information extracted from the original valid data subset, which will have a significant change in the meaning of the original valid data after modification. For example, the original valid data is "As an intelligent voice assistant, you need to act as a navigator to provide users with the best travel route", and the target information to be modified may be "navigator" or "best travel route". If the original valid data is "Please refer to the steps and examples of the above tasks and output the content in json format", the target information to be modified may be "json format". If the original valid data is "Move the co-pilot's chair backwards", the target information to be modified may be "co-pilot" or "chair".

[0136] Target error information refers to the information obtained after the transformation process, which is the error sample that the model should refuse to process. For example, when the target information to be transformed is "navigator", the corresponding target error information may be "psychologist", "shopping guide" and "educator", etc. When the target information to be transformed is "best travel route", the corresponding target error information may be "worst travel route" and "best shopping route", etc. When the target information to be transformed is "json format", the corresponding target error information may be "JavaScript format" and "list format", etc. When the target information to be transformed is "co-pilot", the corresponding target error information may be "window" and "air conditioner", etc. When the target information to be transformed is "chair", the corresponding target error information may be "mobile phone" and "handbag", etc.

[0137] The server parses the original valid data in the original valid data subset to obtain target information to be transformed. Then, the server performs transformation processing on the target information to be transformed to obtain target error information.

[0138] Finally, the server obtains the second training data set based on the target error information. That is, the server uses the target error information to replace the target information to be transformed and generates new training data. For example, using "the worst play route" to replace "the best play route", the original valid data becomes "as an intelligent voice assistant, you need to act as a navigator and provide users with the worst play route".

[0139] In this way, by using the generated target error information to replace the target information to be transformed, a second training data set is generated, so that the model can be trained to better understand under what circumstances it should refuse to process the input, thereby improving the safety and reliability of the model in practical applications.

[0140] See also Figure 7 In some implementations, step 02 (obtaining a third training data set based on the original training data set and the original log data set) includes:

[0141] 021: Based on the preset similarity algorithm, the similarity of each log data is obtained according to the log data in the original log data set and the original valid data subset in the original training data set;

[0142] 022: Determine a third training data set according to a preset similarity threshold and similarity.

[0143] In some embodiments, the processing module is used to obtain the similarity of each log data based on the log data in the original log data set and the original valid data subset in the original training data set based on a preset similarity algorithm, and determine the third training data set based on the preset similarity threshold and the similarity.

[0144] In some embodiments, the processor is further configured to obtain the similarity of each log data according to the log data in the original log data set and the original valid data subset in the original training data set based on a preset similarity algorithm, and determine the third training data set according to the preset similarity threshold and the similarity.

[0145] Specifically, the preset similarity algorithm refers to an algorithm used to calculate the similarity between the log data and the original valid data in the original valid data subset, including TF-IDF weighting (Term Frequency-Inverse Document Frequency), cosine similarity, and Jaccard similarity, etc. For ease of explanation, the implementation of this application takes TF-IDF weighting as an example.

[0146] The original log dataset refers to the user voice request records recorded by the vehicle system, including various inputs encountered by the model in actual applications.

[0147] The preset similarity threshold refers to a similarity critical value set in advance when performing similarity calculation. When the similarity between two texts exceeds this threshold, the two texts can be considered similar. If the similarity is lower than this threshold, they are considered dissimilar. In the implementation of the present application, 0.5 is used as the preset similarity threshold. In some implementations, other values ​​can also be used as the preset similarity threshold, which is not limited here.

[0148] Based on the preset similarity algorithm, the server obtains the similarity of each log data according to the log data in the original log data set and the original valid data subset in the original training data set. In this way, each log data will get a similarity value, indicating its similarity with the known valid samples in the original training data.

[0149] Next, the server determines a third training data set based on a preset similarity threshold and similarity, which is used to determine whether the log data is sufficiently similar to the data in the original valid data subset. If the similarity of the log data is lower than this threshold, it may be regarded as a dissimilar or abnormal input, and it is assembled with a predetermined prompt word and then stored in the third training data set.

[0150] For example, suppose we have an original log data set containing user voice requests and an original training data set containing known valid voice request samples. The server uses a similarity algorithm to calculate the similarity between each log data and the original valid data subset. Suppose we set a similarity threshold of 0.5. For each log data, if the calculated similarity value is lower than 0.5, such as 0.3, the server will regard it as a dissimilar or abnormal input. The server then assembles this log data with a predetermined prompt word (for example, "This is not a valid request"). Finally, the assembled data is stored in the third training data set for training the model's rejection ability.

[0151] In this way, log data that is dissimilar or abnormal to the original valid data subset is obtained, and these log data can be used to train the model to improve the model's ability to reject and process inappropriate or abnormal inputs in practical applications.

[0152] See also Figure 8 In some embodiments, the original valid data subset includes a plurality of original valid data, and step 021 (obtaining the similarity of each log data according to the log data in the original log data set and the original valid data subset in the original training data set based on a preset similarity algorithm) includes:

[0153] 0211: Determine a log data vector of the log data according to a preset similarity algorithm and the log data;

[0154] 0212: Determine an original valid data vector of the original valid data according to a preset similarity algorithm and the original valid data;

[0155] 0213: Get the similarity based on the log data vector and the original valid data vector.

[0156] In some embodiments, the processing module is used to determine the log data vector of the log data according to a preset similarity algorithm and the log data, and to determine the original valid data vector of the original valid data according to the preset similarity algorithm and the original valid data, and to obtain similarity according to the log data vector and the original valid data vector.

[0157] In some embodiments, the processor is further configured to determine a log data vector of the log data according to a preset similarity algorithm and the log data, determine an original valid data vector of the original valid data according to the preset similarity algorithm and the original valid data, and obtain similarity according to the log data vector and the original valid data vector.

[0158] Specifically, the server determines the log data vector of the log data according to the preset similarity algorithm and the log data. Next, the server determines the original valid data vector of the original valid data according to the preset similarity algorithm and the original valid data.

[0159] Finally, the server obtains the similarity based on the log data vector and the original valid data vector. This similarity value can be used for subsequent data processing, such as identifying inputs that are dissimilar or abnormal to the training data, so as to train the rejection capability of the model. In some embodiments, the similarity calculation formula is: Among them, V 日志 is the log data vector of the log data, and V1 is the original valid data vector of the original valid data. 日志 V is the dot product of two vectors, which measures their similarity in direction. ||V 日志|| and ||V1|| are the Euclidean norms (i.e., the lengths of the vectors) of the log data vector and the original valid data vector, respectively. Max indicates that for a given log data vector V 日志 , it is necessary to find the most similar original valid data vector in the original valid data and calculate the similarity.

[0160] For example, suppose we have a log data vector V log representing a user's voice request, and we need to calculate its similarity with the original valid data vector V1.

[0161] First, calculate the dot product of Vlog and each V1, as well as their Euclidean norm. Then, calculate the similarity value according to the similarity formula. Finally, select the V1 with the highest similarity as the most similar original valid data vector. If the calculated similarity value is very low, such as 0.2, we can consider this log data vector Vlog to be abnormal or dissimilar to the training data. We can use it to train the rejection ability of the model to improve the model's ability to handle abnormal inputs.

[0162] In this way, the server can evaluate the similarity between the log data and the original valid data, so as to better understand the relationship between the data and provide a basis for subsequent data processing and analysis.

[0163] See also Fig. 9 In some embodiments, step 0211 (determining a log data vector of the log data according to a preset similarity algorithm and the log data) includes:

[0164] 02111: Determine the word vectors in the log data based on the log data;

[0165] 02112: Determine target words based on log data;

[0166] 02113: According to the target word and the original valid data subset, the word frequency-inverse document frequency weight of the target word is obtained;

[0167] 02114: Obtain the log data vector of the log data based on the word vector and the word frequency-inverse document frequency weight.

[0168] In some embodiments, the processing module is used to determine the word vector in the log data based on the log data. And determine the target word based on the log data. The processing module is used to obtain the word frequency-inverse document frequency weight of the target word based on the target word and the original valid data subset. And obtain the log data vector of the log data based on the word vector and the word frequency-inverse document frequency weight.

[0169] In some embodiments, the processor is further configured to determine a word vector in the log data based on the log data. And determine a target word based on the log data. The processor is further configured to obtain a word frequency-inverse document frequency weight of the target word based on the target word and the original valid data subset. And obtain a log data vector of the log data based on the word vector and the word frequency-inverse document frequency weight.

[0170] Specifically, the target word refers to the word in the log data that needs to be processed. For example, if the log data is "turn on the air conditioner", then the target word can be one of "turn on", "open", "air conditioning" and "adjust".

[0171] Term Frequency-Inverse Document Frequency Weight TF-IDF refers to a statistical method used to evaluate the importance of a term to a document set or one of the documents in a corpus. Term Frequency (TF) refers to the frequency of a term (or word) in a document, and Inverse Document Frequency (IDF) is a measure of the rarity of a term (word) in the entire corpus. The TF-IDF weight is the product of term frequency and inverse document frequency, which can highlight those words that appear frequently in a specific document but are less common in the entire corpus. Term Frequency-Inverse Document Frequency Weight usually includes two parts: First, calculate the term frequency (TF) with the formula TF=i / N, where i is the number of times the word appears in the text and N is the total number of words in the text. Second, calculate the inverse document frequency with the formula IDF=log2(n / k), where n is the total number of sentences or articles in the text library and k is the number of sentences or articles in which the word appears. Third, calculate the term frequency-inverse document frequency weight. The term frequency-inverse document frequency weight is the product of TF and IDF. The calculation formula is TF-IDF=TF*IDF, which can represent the importance of each word in the sentence.

[0172] First, use a pre-trained word embedding model such as BERT to convert each word in the log data into a vector. BERT (Bidirectional Encoder Representations from Transformers) is a pre-trained language representation model that can understand the contextual meaning of words in a sentence. For each word in a sentence, BERT generates a fixed-dimensional vector that contains the semantic information of the word.

[0173] Then, based on the log data obtained from the self-log dataset, determine the target word. For example, if the obtained log data is "open the car door", the target word is determined to be "car". Subsequently, use the original valid data subset as the text library, that is, the total number of words in the original valid data subset is N, the total number of "car" in the original valid data subset is i, the total number of rain gear in the original valid data subset is n, and the number of sentences containing "car" in the original valid data subset is k. Calculate the term frequency-inverse document frequency weight TF-IDF of the target word "car". Then, calculate the term frequency-inverse document frequency weight TF-IDF of the word "open" in the target data "open the car door", calculate the term frequency-inverse document frequency weight TF-IDF of the word "door" in the target data "open the car door", and calculate the term frequency-inverse document frequency weight TF-IDF of the word "door" in the target data "open the car door".

[0174] Finally, based on the term frequency-inverse document frequency weights TF-IDF of each word in the log data and the word vectors of each word, perform a weighted sum to obtain the log data vector of the log data. It should be noted that the original valid data vector is obtained based on similar steps and will not be elaborated here.

[0175] In this way, a log data vector composed of the TF-IDF weights of the target words is obtained, which can effectively represent the characteristics and content of the log data, thereby better understanding the content and theme of the data for subsequent similarity calculation.

[0176] Please refer to Fig.10 , in some embodiments, step 03 (training the model according to the original training dataset, the first training dataset, the second training dataset, and the third training dataset to improve the rejection ability of the model) includes:

[0177] 031: Obtain a rejection training dataset according to the second training dataset and the third training dataset;

[0178] 032: Based on a preset ratio, train the model according to the original training dataset, the first training dataset, and the rejection training dataset to improve the rejection ability of the model.

[0179] In some embodiments, the training module is used to obtain a rejection training dataset according to the second training dataset and the third training dataset, and train the model according to the original training dataset, the first training dataset, and the rejection training dataset based on a preset ratio to improve the rejection ability of the model.

[0180] In some embodiments, the processor is used to obtain a rejection training data set based on the second training data set and the third training data set, and train the model based on the original training data set, the first training data set and the rejection training data set based on a preset ratio to improve the rejection capability of the model.

[0181] Specifically, the preset ratio refers to the ratio relationship between the original training data set, the first training data set and the rejection training data set when training the model.

[0182] See also Fig.11 , Fig.11 Schematic diagram of the model training data processing flow. Among them, first, by adding specific rejection task prompts (i.e., preset rejection descriptions) to the original valid data in the original training data, the original valid data with the rejection task prompts added is transformed to meet the needs of the rejection model, thereby obtaining the first training data set. Second, some task information in the original training data is updated, that is, some information (target information to be transformed) in the original valid data that can give a normal answer is transformed. And the updated data is further transformed to better meet the needs of the rejection model to obtain the second training data set. Third, log data is obtained, which may include problems and challenges encountered in practical applications, and the log data is screened according to the original valid data, and sentences with low similarity to the original training data are selected to increase the generalization ability of the model. Subsequently, the screened data is further transformed to better meet the needs of the rejection model to obtain the third training data set. Fourth, the original training data set, the first training data set, the second training data set and the third training data set are used to enhance the rejection ability of the model. Figure 1 It shows the process of starting from the original training data, adding rejection tasks, updating data, screening and transformation, and finally training a rejection model based on the obtained training data.

[0183] In this way, the server determines the usage of data in the original training data set, the first training data set and the rejection training data set based on the preset ratio to achieve good training results, thereby improving the model's ability to refuse to process inappropriate or abnormal inputs in practical applications.

[0184] See also Fig.12 The embodiment of the present application provides a voice interaction method, which is based on a model optimized by the above-mentioned optimization method, and the method includes:

[0185] 04: Receive the voice request forwarded by the vehicle;

[0186] 05: Reject the voice request based on the model to complete the voice interaction.

[0187] The embodiment of the present application also provides a server, including a memory and a processor. The voice interaction method of the embodiment of the present application can be implemented by the server of the embodiment of the present application. Specifically, a computer program is stored in the memory, and the processor is used to receive a voice request forwarded by the vehicle. And the voice request is rejected according to the model to complete the voice interaction.

[0188] The embodiment of the present application also provides a voice interaction device. The voice interaction method of the embodiment of the present application can be implemented by the voice interaction device of the embodiment of the present application. Specifically, the voice interaction device includes a receiving module and a rejection module. The receiving module is used to receive the voice request forwarded by the vehicle. The rejection module is used to reject the voice request according to the model to complete the voice interaction.

[0189] Specifically, the server receives the voice request forwarded by the vehicle. Then, the server analyzes the received voice request according to the model rejection model obtained above to complete the voice interaction. The task of the rejection model is to determine whether the voice request is suitable or can be processed by the system. If the request is irrelevant to the function of the system or cannot be understood, the model will recognize and reject the voice request.

[0190] For example, suppose a user in a vehicle says, "Please help me find the nearest gas station." The voice request is forwarded by the vehicle to the server. The server analyzes the request using a rejection model. The model recognizes that this is a valid request because it is related to the navigation function of the vehicle system. The server then processes the request and returns navigation information to the nearest gas station to the vehicle.

[0191] If the user says, “Please tell me a joke.”, the rejection model may recognize that this is an inappropriate request for the vehicle system to handle because it is not related to the vehicle’s functions (such as navigation, music playback, etc.). Therefore, the model will reject the request and may prompt the user to say, “Sorry, I can’t perform this request.”

[0192] In summary, in the voice interaction method and server provided in the embodiment of the present application, the server receives the voice request forwarded by the vehicle. Then, the server rejects the voice request according to the model to complete the voice interaction. In this way, based on the model trained by the above optimization method, the server can ensure that the voice interaction between the vehicle and the user is both safe and reliable.

[0193] The present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned optimization method or voice interaction method are implemented.

[0194] It is understood that a computer program includes computer program code. The computer program code may be in source code form, object code form, executable file or some intermediate form. Computer readable storage media may include: any entity or device capable of carrying computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), and software distribution medium.

[0195] In the description of this specification, the descriptions with reference to the terms "specifically", "further", "particularly", "understandably", etc. are intended to mean that the specific features, structures, materials or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms are not intended to refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0196] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code that includes one or more executable requests for implementing specific logical functions or steps of a process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0197] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for optimizing a model for vehicle voice interaction, characterized in that: The method comprises: Performing rejection transformation processing on the original training data set of the model to obtain a first training data set and a second training data set; Obtaining a third training data set according to the original training data set and the original log data set; The model is trained according to the original training data set, the first training data set, the second training data set and the third training data set to improve the rejection capability of the model.

2. The optimization method according to claim 1, characterized in that: The reject-recognition transformation process is performed on the original training data set of the model to obtain the first training data set and the second training data set, including: Based on a preset test method, the first training data set is obtained according to a preset rejection description, an original rejection data subset and an original valid data subset in the original training data set.

3. The optimization method according to claim 2, characterized in that: The original valid data subset includes a plurality of original valid data, and the first training data set is obtained based on the preset test method according to the original training data set and the preset rejection description, including: Based on a preset granularity, segment each of the original valid data in the original valid data subset to obtain a plurality of target rejected data, wherein each of the target rejected data corresponds to one of the original valid data; According to the preset rejection description and the plurality of target rejection data, a plurality of temporary rejection data sets are obtained; The temporary rejected data set is tested according to the original rejected data subset to confirm that at least one of the temporary rejected data sets is added to the first training data set.

4. The optimization method according to claim 3, characterized in that: The target rejection data includes at least one sentence, and the multiple temporary rejection data sets are obtained according to the preset rejection description and the target rejection data, including: The preset rejection description is placed at a segmentation position in each of the target rejection data to obtain a plurality of temporary rejection data sets, wherein the segmentation position is a connection position of each sentence in the target rejection data.

5. The optimization method according to claim 2, characterized in that: The reject-recognition transformation process is performed on the original training data set of the model to obtain the first training data set and the second training data set, including: The original valid data subset is transformed to obtain the second training data set.

6. The optimization method according to claim 5, characterized in that: The step of transforming the original valid data subset to obtain the second training data set includes: Analyzing and processing the original valid data in the original valid data subset to obtain target information to be transformed; Performing transformation processing on the target information to be transformed to obtain target error information; The second training data set is obtained according to the target error information.

7. The optimization method according to claim 1, characterized in that: The step of obtaining a third training data set according to the original training data set and the original log data set includes: Based on a preset similarity algorithm, obtaining the similarity of each log data according to the log data in the original log data set and the original valid data subset in the original training data set; The third training data set is determined according to a preset similarity threshold and the similarity.

8. The optimization method according to claim 7, characterized in that: The original valid data subset includes a plurality of original valid data, and the similarity of each log data is obtained based on the log data in the original log data set and the original valid data subset in the original training data set based on the preset similarity algorithm, including: Determining a log data vector of the log data according to the preset similarity algorithm and the log data; Determining an original valid data vector of the original valid data according to the preset similarity algorithm and the original valid data; The similarity is obtained according to the log data vector and the original valid data vector.

9. The optimization method according to claim 8, characterized in that: The step of determining the log data vector of the log data according to the preset similarity algorithm and the log data includes: Determine, based on the log data, a word vector in the log data; Determine a target word according to the log data; According to the target word and the original valid data subset, a word frequency-inverse document frequency weight of the target word is obtained; A log data vector of the log data is obtained according to the word vector and the word frequency-inverse document frequency weight.

10. The optimization method according to claim 1, characterized in that: The training of the model according to the original training data set, the first training data set, the second training data set and the third training data set to improve the rejection capability of the model includes: Obtaining a rejection training data set according to the second training data set and the third training data set; Based on a preset ratio, the model is trained according to the original training data set, the first training data set and the rejection training data set to improve the rejection capability of the model.

11. A voice interaction method, characterized in that: The voice interaction method is based on a model optimized by the optimization method according to any one of claims 1 to 10, and the method comprises: Receive voice requests forwarded by the vehicle; The voice request is rejected according to the model to complete the voice interaction.

12. A server, characterized in that: The server includes a processor and a memory, wherein a computer program is stored in the memory. When the computer program is executed by the processor, the optimization method described in any one of claims 1 to 11 is implemented.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

Images