Intelligent customer service automatic recognition language reply method and device, computer equipment, readable storage medium and product

Abstract

The invention relates to an intelligent customer service automatic recognition language reply method and device, computer equipment, a computer readable storage medium and a computer program product, and relates to the technical field of data processing. The method comprises the following steps: receiving consultation content initiated by a user through a user terminal; the consultation content comprises voice consultation content and / or text consultation content; identifying the consultation content by adopting a language model and / or an acoustic model to obtain language features corresponding to the consultation content; the language feature represents a consultation intention corresponding to the consultation content; according to the language features, a neighbor interpolation algorithm is adopted to retrieve preset reply data, and under the condition that a target preset reply is retrieved, the target preset reply is sent to the user terminal; the preset reply data is obtained by integrating the historical consultation content and the historical reply data. By adopting the method, the automatic reply accuracy of the intelligent customer service can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, device, computer equipment, computer-readable storage medium and computer program product for automatically identifying language responses by intelligent customer service. Background Art

[0002] In the case where the current user initiates customer service consultation, the soft switch intelligent voice interaction system that usually uses intelligent voice technology can replace the mobile communication operator customer service to conduct multiple rounds of dialogue with the user. Among them, the soft switch intelligent voice interaction system includes a soft switch subsystem and an intelligent voice subsystem. After receiving the call voice from the user, the mobile communication operator sends the call voice to the soft switch subsystem, so that the soft switch subsystem sends the call voice to the intelligent voice subsystem. After the intelligent voice subsystem generates the corresponding reply voice, it sends the reply voice to the soft switch subsystem, and the soft switch subsystem sends the reply voice to the mobile communication operator, and then the mobile communication operator feeds back the reply voice to the customer, thereby completing this round of dialogue.

[0003] However, the current intelligent response system cannot intelligently integrate the questions asked by each customer, resulting in the same mechanical answers given when customers ask the same questions, and the accuracy of the responses given is low. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for automatically identifying language responses of intelligent customer service, which can improve the accuracy of automatic responses of intelligent customer service, in response to the above technical problems.

[0005] In a first aspect, the present application provides a method for automatically identifying language responses by intelligent customer service, comprising:

[0006] Receiving consultation content initiated by a user through a user terminal; the consultation content includes voice consultation content and / or text consultation content;

[0007] The consultation content is identified by using a language model and / or an acoustic model to obtain language features corresponding to the consultation content; the language features represent the consultation intention corresponding to the consultation content;

[0008] According to the language features, a neighbor interpolation algorithm is used to retrieve the preset reply data. When the target preset reply is retrieved, the target preset reply is sent to the user terminal; the preset reply data is obtained based on the integration of historical consultation content and historical reply data.

[0009] In one embodiment, the use of a language model or an acoustic model to identify the consultation content to obtain language features corresponding to the consultation content includes:

[0010] In the case where the consultation content includes voice consultation content, the consultation content is processed using an acoustic model to obtain consultation voice data;

[0011] Using a language model to recognize the consultation voice data to obtain language features;

[0012] In the case where the consultation content includes text consultation content, a language model is used to identify the text consultation content to obtain language features.

[0013] In one embodiment, the retrieving the preset reply data using a neighbor interpolation algorithm according to the language feature includes:

[0014] Determining the characteristic words included in the language feature;

[0015] In the case where the characteristic word is a positive characteristic word, a neighbor interpolation algorithm is used to retrieve a target positive characteristic word from a preset characteristic word library;

[0016] According to the target positive characteristic word and the first mapping relationship, the preset reply data is retrieved to obtain the target preset reply.

[0017] In one embodiment, the method described in the above embodiment further includes:

[0018] In the case where the characteristic word is a negative characteristic word, a neighbor interpolation algorithm is used to retrieve a target negative characteristic word from a preset characteristic word library;

[0019] According to the target negative characteristic word and the second mapping relationship, the preset reply data is retrieved to obtain the target preset reply.

[0020] In one embodiment, before searching the preset reply data using the neighbor interpolation algorithm according to the language feature, the method further includes:

[0021] Constructing a first mapping relationship according to the positive characteristic words included in the preset characteristic word library and the preset reply data corresponding to the positive characteristic words;

[0022] Constructing a second mapping relationship according to the negative characteristic words included in the preset characteristic word library and the preset reply data corresponding to the negative characteristic words;

[0023] A third mapping relationship is constructed based on all the preset reply data and the voice replies corresponding to the preset reply data; the third mapping relationship is used to send the voice reply corresponding to the target preset reply to the user terminal when the consultation content includes voice consultation content.

[0024] In one embodiment, the method described in the above embodiment further includes:

[0025] In the case where the target preset reply is not retrieved, calling the manual customer service interface according to the peak intelligent queuing algorithm to connect the communication between the user terminal and the manual customer service;

[0026] The reply data of the manual customer service after the communication is connected is obtained through the manual customer service interface, the reply data of the manual customer service interface is associated with the language feature, and the preset reply data is updated.

[0027] In a second aspect, the present application also provides an intelligent customer service automatic recognition language reply device, comprising:

[0028] A consultation receiving module, used to receive consultation content initiated by a user through a user terminal; the consultation content includes voice consultation content and / or text consultation content;

[0029] A feature recognition module is used to recognize the consulting content using a language model and / or an acoustic model to obtain language features corresponding to the consulting content; the language features represent the consulting intention corresponding to the consulting content;

[0030] The consultation reply module is used to retrieve the preset reply data according to the language features using a neighbor interpolation algorithm, and when the target preset reply is retrieved, the target preset reply is sent to the user terminal; the preset reply data is obtained based on the integration of historical consultation content and historical reply data.

[0031] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0032] Receiving consultation content initiated by a user through a user terminal; the consultation content includes voice consultation content and / or text consultation content;

[0033] The consultation content is identified by using a language model and / or an acoustic model to obtain language features corresponding to the consultation content; the language features represent the consultation intention corresponding to the consultation content;

[0034] According to the language features, a neighbor interpolation algorithm is used to retrieve the preset reply data. When the target preset reply is retrieved, the target preset reply is sent to the user terminal; the preset reply data is obtained based on the integration of historical consultation content and historical reply data.

[0035] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0036] Receiving consultation content initiated by a user through a user terminal; the consultation content includes voice consultation content and / or text consultation content;

[0037] The consultation content is identified by using a language model and / or an acoustic model to obtain language features corresponding to the consultation content; the language features represent the consultation intention corresponding to the consultation content;

[0038] According to the language features, a neighbor interpolation algorithm is used to retrieve the preset reply data. When the target preset reply is retrieved, the target preset reply is sent to the user terminal; the preset reply data is obtained based on the integration of historical consultation content and historical reply data.

[0039] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0040] Receiving consultation content initiated by a user through a user terminal; the consultation content includes voice consultation content and / or text consultation content;

[0041] The consultation content is identified by using a language model and / or an acoustic model to obtain language features corresponding to the consultation content; the language features represent the consultation intention corresponding to the consultation content;

[0042] According to the language features, a neighbor interpolation algorithm is used to retrieve the preset reply data. When the target preset reply is retrieved, the target preset reply is sent to the user terminal; the preset reply data is obtained based on the integration of historical consultation content and historical reply data.

[0043] The above-mentioned intelligent customer service automatic recognition language reply method, device, computer equipment, computer-readable storage medium and computer program product receive the consultation content initiated by the user through the user terminal, wherein the consultation content includes voice consultation content and / or text consultation content, so that the services that the intelligent customer service can respond to include both voice and text. The consultation content is identified by a language model and / or an acoustic model to obtain the language features corresponding to the consultation content, wherein the language features characterize the consultation intention corresponding to the consultation content. The user's consultation intention is identified by the artificial intelligence model, which can improve the recognition accuracy and the recognition efficiency. Further, according to the language features, the neighbor interpolation algorithm is used to retrieve the preset reply data, and when the target preset reply is retrieved, the target preset reply is sent to the user terminal; the preset reply data is obtained based on the integration of historical consultation content and historical reply data. The preset reply data can be continuously integrated and updated as the user consultation tasks increase, thereby enhancing the ability and accuracy of the automatic reply of the intelligent customer service. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0045] FIG1 is a diagram showing an application environment of a method for automatically identifying language responses by intelligent customer service in one embodiment;

[0046] FIG2 is a schematic diagram of a flow chart of a method for automatically identifying language responses by intelligent customer service in one embodiment;

[0047] FIG3 is a schematic diagram of a process of constructing a mapping relationship in one embodiment;

[0048] FIG4 is a flow chart of a method for automatically identifying language responses by intelligent customer service in another embodiment;

[0049] FIG5 is a structural block diagram of an intelligent customer service automatic recognition language reply device in one embodiment;

[0050] FIG. 6 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0052] The method for automatically identifying language responses by intelligent customer service provided in the embodiment of the present application can be applied in the application environment shown in FIG1. ​​The application scenario includes: a user terminal 102, a server 104, and a data storage system. Among them, the user terminal 102 communicates with the server 104 via a network or satellite communication. The data storage system can store data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices. The portable wearable device can be a smart watch or a smart bracelet. The server 104 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The server 104 receives the consulting content initiated by the user through the user terminal 102, wherein the consulting content includes voice consulting content and / or text consulting content. The server 104 uses a language model and / or an acoustic model to identify the consulting content, and obtains the language features corresponding to the consulting content, wherein the language features represent the consulting intention corresponding to the consulting content. The server 104 further uses a neighbor interpolation algorithm to retrieve preset reply data based on language features, and when a target preset reply is retrieved, the target preset reply is sent to the user terminal, wherein the preset reply data is obtained based on the integration of historical consultation content and historical reply data.

[0053] In an exemplary embodiment, as shown in FIG2 , a method for automatically identifying language responses by intelligent customer service is provided, which is described by taking the method applied to the server 104 in FIG1 as an example, and includes the following steps S202 to S206 . Among them:

[0054] Step S202: receiving the consultation content initiated by the user through the user terminal.

[0055] The consultation content includes voice consultation content and / or text consultation content. The content input by the user can be in the form of text from the official website, WeChat public account or mini program, or voice call from the call center, etc.

[0056] Optionally, the server receives voice consultation content included in the consultation content initiated by the user through the user terminal in the form of a telephone call, or receives text consultation content included in the consultation content typed by the user into the customer service consultation input box.

[0057] Step S204: Use a language model and / or an acoustic model to identify the consultation content to obtain language features corresponding to the consultation content.

[0058] Among them, language features represent the consulting intention corresponding to the consulting content.

[0059] Among them, the semantic model is a core part of the speech recognition system, which is responsible for converting speech features into text information. These text information can be words, sentences or higher-level semantic information. The acoustic model is another core part of the speech recognition system, which is responsible for converting speech signals into speech features. These features include spectral features, time domain features, and time-frequency features.

[0060] Optionally, the server uses a language model to identify text consultation content included in the consultation content, and the server uses an acoustic model and a language model to identify voice consultation content included in the consultation content to obtain language features corresponding to the consultation content.

[0061] Step S206: According to the language features, a neighbor interpolation algorithm is used to retrieve the preset reply data. When the target preset reply is retrieved, the target preset reply is sent to the user terminal.

[0062] Among them, the preset response data is obtained based on the integration of historical consultation content and historical response data.

[0063] Among them, the neighbor interpolation algorithm, also known as the nearest neighbor fitting interpolation, is a simple multivariate interpolation method. Its basic idea is to select the nearest data point around the interpolation point and use this point as the support for interpolation.

[0064] Optionally, the server uses a neighbor interpolation algorithm to determine the identifier that is closest to the language feature based on the language feature, and searches based on the identifier. When a target preset reply corresponding to the identifier is retrieved, the target preset reply is used as the reply corresponding to the consultation content, and the target preset reply is sent to the user terminal in the form of text or voice.

[0065] In the above-mentioned method for automatically identifying language replies by intelligent customer service, by receiving the consultation content initiated by the user through the user terminal, wherein the consultation content includes voice consultation content and / or text consultation content, the services that the intelligent customer service can respond to include both voice and text. The consultation content is identified by a language model and / or an acoustic model to obtain the language features corresponding to the consultation content, wherein the language features characterize the consultation intention corresponding to the consultation content. By using an artificial intelligence model to identify the user's consultation intention for the user's consultation content, it is possible to improve the recognition accuracy while also improving the recognition efficiency. Further, based on the language features, a neighbor interpolation algorithm is used to retrieve the preset reply data, and when the target preset reply is retrieved, the target preset reply is sent to the user terminal; the preset reply data is obtained based on the integration of historical consultation content and historical reply data. The preset reply data can be continuously integrated and updated as the user consultation tasks increase, thereby enhancing the ability and accuracy of the automatic reply of the intelligent customer service.

[0066] In an exemplary embodiment, step S204 uses a language model or an acoustic model to identify the consultation content to obtain language features corresponding to the consultation content, including:

[0067] When the consultation content includes voice consultation content, the acoustic model is used to process the consultation content to obtain consultation voice data; the language model is used to recognize the consultation text data to obtain language features; when the consultation content includes text consultation content, the language model is used to recognize the text consultation content to obtain language features.

[0068] The consultation voice data may be a probability distribution result of converting an audio signal into phonemes or feature vectors after the acoustic model processes the voice consultation content.

[0069] Optionally, when the consultation content includes voice consultation content, automatic Yuying recognition technology and voice understanding are used to identify whether the consultation content initiated by the user is repetitive. If it is not repetitive, the consultation content is processed using an acoustic model to obtain consultation voice data, including the probability distribution results of corresponding phonemes or feature vectors. The consultation voice data is recognized using a language model, such as by using a language model to distribute the probability of phonemes and feature vectors and understand the contextual relationship to obtain language features. When the consultation content includes text consultation content, the text consultation content is recognized using a language model to obtain language features.

[0070] In this embodiment, combining the acoustic model with the language model not only enhances the accuracy and robustness of speech recognition, but also improves the ability to process multimodal content, laying the foundation for more complex natural language understanding for intelligent customer service.

[0071] In an exemplary embodiment, step S206 uses a neighbor interpolation algorithm to retrieve the preset reply data according to the language features, including:

[0072] Determine the characteristic words included in the language feature; when the characteristic words are affirmative characteristic words, use a neighbor interpolation algorithm to retrieve the target affirmative characteristic words from a preset characteristic word library; based on the target affirmative characteristic words and the first mapping relationship, retrieve the preset reply data to obtain the target preset reply.

[0073] Among them, characteristic words may refer to words used in linguistics to describe language features or properties. They are usually used to describe the form, function or meaning of words or sentences. In this embodiment, it refers to words with positive or negative characteristics. Among them, positive characteristic words may be words used to express affirmation, approval or confirmation. The preset characteristic word library may be constructed in advance based on multiple characteristic words for which corresponding reply data exists.

[0074] The first mapping relationship may be an association relationship between positive characteristic words included in a preset characteristic word library and their corresponding preset replies. The preset reply data may be a set of preset replies with an association relationship.

[0075] Optionally, the server determines the characteristic words included in the language feature and determines the type of the characteristic words. When the characteristic words are positive characteristic words, a neighbor interpolation algorithm is used to retrieve the target positive characteristic words that are closest to the characteristic words of the language feature from a preset characteristic word library. Further, based on the target positive characteristic words and the first mapping relationship, a search is performed from the preset reply data to retrieve a preset reply corresponding to the target positive characteristic word as the target preset reply corresponding to the language feature.

[0076] In this embodiment, through the analysis of characteristic words and the application of the neighbor interpolation algorithm, the server of the intelligent customer service can obtain more relevant and personalized replies. The neighbor interpolation algorithm enables the system to find the most suitable target positive characteristic words in the diverse consultation content, and further retrieve the most suitable target preset reply, thereby improving the accuracy of the automatic reply of the intelligent customer service.

[0077] In an exemplary embodiment, the method described in the above embodiment further includes:

[0078] When the characteristic word is a negative characteristic word, a neighbor interpolation algorithm is used to retrieve the target negative characteristic word from the preset characteristic word library; based on the target negative characteristic word and the second mapping relationship, the preset reply data is retrieved to obtain the target preset reply.

[0079] Among them, negative characteristic words can be words used to express negation, opposition or rejection.

[0080] The second mapping relationship may be an association relationship between a plurality of preset negative characteristic words in a preset characteristic word library and corresponding preset replies.

[0081] Optionally, when the characteristic word is a negative characteristic word, the server uses a neighbor interpolation algorithm to retrieve a target negative characteristic word that is closest to the characteristic word corresponding to the language feature from a preset characteristic word library. And according to the target negative characteristic word and the second mapping relationship, the preset reply data is retrieved to retrieve a preset reply corresponding to the target negative characteristic word, and the preset reply is used as the target preset reply corresponding to the language feature.

[0082] In this embodiment, the recognition of positive and negative characteristic words can identify the user's emotions, and the intelligent customer service server can retrieve more relevant and personalized replies, further improving the accuracy of the intelligent customer service automatic reply.

[0083] In an exemplary embodiment, as shown in FIG3 , before step S206 uses a neighbor interpolation algorithm to search for the preset reply data according to the language features, the following steps S302 to S306 are also included. Among them:

[0084] Step S302: construct a first mapping relationship according to the positive characteristic words included in the preset characteristic word library and the preset reply data corresponding to the positive characteristic words.

[0085] Optionally, the server constructs a first mapping relationship based on the association between the positive characteristic words included in the preset characteristic word library and the preset response data corresponding to each positive characteristic word, for example, by taking each positive characteristic word as an identifier and the corresponding preset response data as the content corresponding to the identifier to construct a mapping table.

[0086] Step S304: construct a second mapping relationship according to the negative characteristic words included in the preset characteristic word library and the preset reply data corresponding to the negative characteristic words.

[0087] Optionally, the server constructs a second mapping relationship based on the association between the positive characteristic words included in the preset characteristic word library and the preset response data corresponding to each negative characteristic word. For example, a mapping table is constructed by taking each negative characteristic word as an identifier and the corresponding preset response data as the content corresponding to the identifier.

[0088] Step S306: construct a third mapping relationship according to all preset reply data and the voice reply corresponding to the preset reply data.

[0089] Among them, the third mapping relationship is used to send the voice reply corresponding to the target preset reply to the user terminal when the consultation content includes voice consultation content.

[0090] The voice reply may be generated by using speech synthesis technology according to preset reply data.

[0091] Optionally, the server constructs a third mapping relationship based on the association between all preset reply data and the voice reply of each reply data, for example, by taking each preset reply data as an identifier and the corresponding voice reply as the content corresponding to the identifier to construct a mapping table.

[0092] In this embodiment, by constructing three mapping relationships, response data for different consultation situations are stored in advance, and the intelligent customer service can respond to the user's consultation request in a timely and accurate manner.

[0093] In an exemplary embodiment, the method described in the above embodiment further includes:

[0094] If the target preset reply is not retrieved, the manual customer service interface is called according to the peak period intelligent queuing algorithm to connect the communication between the user terminal and the manual customer service; through the manual customer service interface, the reply data of the manual customer service after the communication is connected is obtained, the reply data of the manual customer service interface is associated with the language features, and the preset reply data is updated.

[0095] Among them, the peak period intelligent queuing algorithm can be an algorithm for managing high load situations in voice or text consultation services.

[0096] Optionally, if the target preset reply is not retrieved, it indicates that the intelligent customer service cannot solve the consultation content initiated by the user at this time. The server calls the manual customer service interface according to the peak intelligent queuing algorithm to connect the communication between the user terminal and the manual customer service. The peak intelligent queuing algorithm can be processed according to the first-come-first-served principle and the order of user requests. Different queuing strategies can also be adopted to better manage user waiting time and service quality. For example, priority queuing can be implemented to determine the service order of users according to user type, problem type or other priority criteria. It should be noted that the goal of the peak intelligent queuing algorithm is to maintain a good service level during peak hours, and to ensure that users receive timely service responses during the waiting process through reasonable queuing management and transfer to manual customer service. This algorithm can help the robot system better cope with high load conditions and improve the overall service quality and user experience. The server obtains the reply data of the manual customer service after the communication is connected through the manual customer service interface, associates the reply data of the manual customer service interface with the language features, and updates the preset reply data.

[0097] In this embodiment, when the target preset reply cannot be retrieved, the call is transferred to the manual customer service, and the user is queued through the peak intelligent queuing algorithm, which can help the robot system better cope with high load conditions and improve the overall service quality and user experience. In addition, the reply data of the manual customer service is obtained to update the preset reply data, thereby optimizing the intelligent customer service and improving the accuracy of the intelligent customer service's automatic reply in subsequent consultations.

[0098] In an exemplary embodiment, as shown in FIG4 , another method for automatically identifying language responses by intelligent customer service is provided. The method includes:

[0099] S1: Start.

[0100] S2: Receive customer service consultation. Among them, the voice consultation is identified according to the acoustic and language features of the voice consultation through the voice recognition model. Among them, the text content of the voice consultation can be obtained through the voice recognition model to facilitate subsequent answers. Among them, the acoustic features may include energy features, time domain features, frequency domain features, music theory features and perceptual features. The time domain feature value is the autocorrelation value, wherein the autocorrelation value refers to the similarity between the signal and its version after displacement along the time axis, which can be used to calculate the fundamental frequency of a single tone. The frequency domain feature value is the spectrum centroid value, wherein the spectrum centroid value refers to the concentration point of the signal energy in the spectrum, which can be used to describe the brightness of the signal timbre. It can be understood that the brighter the sound energy is concentrated in the high frequency part, the greater the value of the spectrum centroid; the music theory feature value is the detuning value, wherein the detuning value refers to the degree of deviation between the overtone frequency of the audio signal and the integer multiple of its fundamental frequency; the fundamental frequency, referred to as the fundamental frequency, the sound can be decomposed into the superposition of several sine waves of different frequencies, the wave with the lowest frequency is the fundamental tone, and the other high frequencies are overtones, and the higher the frequency, the less energy is allocated. The perceptual feature value is the loudness value, where the loudness value refers to the subjective feeling of the signal strength felt by the human ear, which can also be understood as the volume. Among them, language features may include: semantic features, lexical features, grammatical features, etc. Semantic features are the meanings of vocabulary and sentences. Lexical features refer to the composition, meaning, word form changes, etc. of lexical features. Grammatical features refer to the structural rules of language. It should be noted that through the speech recognition model, the voice consultation is identified according to the acoustic features and language features of the voice consultation, which greatly improves the accuracy of speech recognition compared to the single acoustic feature. In one possible implementation, the speech recognition model includes an acoustic model and a language model. The role of the acoustic model is to calculate the probability that the acoustic feature sequence belongs to each acoustic unit, and convert the feature sequence into a state sequence according to the likelihood criterion. The language model describes the constraints between words in linguistics, and is used to predict the probability that the next word may appear in the speech signal.

[0101] S3: Obtain the content of the customer's consultation (corresponding to the consultation content in the above embodiment), and determine the user's intention based on the consultation content. Among them, obtain the content input by the user, and determine the user's intention based on the content. Among them, the content input by the user can be in the form of text from the official website, WeChat public account or mini program, or it can be a voice call from a call center, etc. In one embodiment, before step S3, it also includes: building a product knowledge base for the product to be consulted. Among them, the product knowledge base may include but is not limited to the following elements: product name, product dimension, segmentation attribute, attribute value, product pricing strategy, product maintenance strategy, common fault diagnosis and repair steps.

[0102] S4: The reply data most relevant to the consultation (corresponding to the target preset reply in the above embodiment) is searched through the nearest neighbor interpolation algorithm (corresponding to the neighbor interpolation algorithm in the above embodiment), and the robot automatically replies to the customer according to the reply data. Among them, the reply data most relevant to the voice consultation is searched through the nearest neighbor query algorithm, and the robot automatically replies to the customer according to the reply data. Among them, the nearest neighbor query algorithm is an algorithm for finding the most similar data points in a data set. In the context of voice consultation, this data set contains a series of reply data, each reply is associated with a specific consultation question or voice input. The nearest neighbor query algorithm can find the most similar reply data according to the characteristics of the consultation question or voice input. It should be noted that once the reply data most relevant to the consultation question or voice input is found, the robot can automatically extract the corresponding reply content and send it to the customer as an automatic reply. This process can be achieved through natural language processing technology and speech synthesis technology to ensure that the generated reply content is accurate, fluent and matches the user's consultation question. Therefore, the efficiency of customer service can be improved, the need for manual intervention can be reduced, and customers can be provided with fast and accurate replies under certain similarity matching conditions.

[0103] S5: Identify whether the customer is hard of hearing through automatic speech recognition technology and speech understanding technology. When the user is not hard of hearing, obtain the user's voice in the interjection dialogue between the user and the customer service, and use automatic speech recognition technology to convert the obtained user's voice into text; analyze the converted text through language understanding technology to obtain text information in the text that represents the user's intention; match the question answer corresponding to the text information in the preset knowledge base based on the text information; and reply the matched question answer to the user.

[0104] S6: When the robot is unable to solve the problem being consulted, you can choose to transfer to manual service through the peak queuing algorithm. Among them, the peak intelligent queuing algorithm is an algorithm used to manage high load conditions in voice consultation services. When the robot cannot effectively solve the voice consultation problem, this algorithm can automatically transfer the user to the manual customer service and manage a large number of user requests through a queuing mechanism. Regarding the queuing strategy, the peak intelligent queuing algorithm can process the user's request order according to the first-come-first-served principle. Different queuing strategies can also be adopted to better manage user waiting time and service quality. For example, priority queuing can be implemented to determine the service order of users according to user type, problem type or other priority criteria. It should be noted that the goal of the peak intelligent queuing algorithm is to maintain a good service level during peak hours, and ensure that users receive timely service responses during the waiting process through reasonable queuing management and transfer to manual customer service. This algorithm can help the robot system better cope with high load conditions and improve the overall service quality and user experience.

[0105] Step S6 includes the following steps:

[0106] S61: The manual customer service answers the questions asked by the customer; if the manual customer service is unable to solve the voice consultation, a work order is created and the work order content is processed for privacy protection. The work order is used to arrange for professionals to answer the voice consultation questions. It should be noted that the privacy protection of the work order content can protect customer privacy, build customer trust, prevent information abuse and anonymize data.

[0107] S62: Intelligently integrate the questions asked by the customer. The questions asked by the customer are integrated again. In the process of repeated integration, the intelligent response becomes clearer and more intelligent, and the customer's problem is solved as quickly as possible.

[0108] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0109] Exemplarily, in this embodiment, the consultation content in step S62 is stored and calculated. The consultation content in step S62 is stored and calculated, including a receiving unit for receiving the voice to be replied.

[0110] A judgment unit is used to judge whether the text content corresponding to the voice to be replied is a target specific word based on a target automatic speech recognition module, wherein the target automatic speech recognition module is a functional module pre-integrated in the soft switch subsystem, and the target specific word (corresponding to the preset characteristic word library of the above embodiment) is any one of a plurality of preset positive words (corresponding to the positive characteristic words of the above embodiment) and a plurality of preset negative words (corresponding to the negative characteristic words of the above embodiment).

[0111] A generating unit is used for generating a reply voice (corresponding to the voice reply in the above-mentioned embodiment) corresponding to the voice to be replied according to the target specific word, the target natural language understanding module and the target text-to-speech module when the judging unit determines that the text content corresponding to the voice to be replied is the target specific word, wherein the target natural language understanding module and the target text-to-speech module are functional modules pre-integrated in the soft switch subsystem.

[0112] The sending unit is used to send the to-be-replied voice to the intelligent voice subsystem when the judgment unit determines that the text content corresponding to the to-be-replied voice is not the target specific word, so that the intelligent voice subsystem generates a reply voice corresponding to the to-be-replied voice. The target automatic speech recognition module includes a preset acoustic model, a preset language model and a preset recognition word network.

[0113] The judgment unit includes: a first input module, which is used to input the speech to be replied into the target automatic speech recognition module, so that the target automatic speech recognition module performs speech recognition on the speech to be replied based on a preset acoustic model, a preset language model and a preset recognition word network.

[0114] The first determination module is used to determine that the text content corresponding to the to-be-answered speech is the target specific word when the target automatic speech recognition module outputs the target specific word; the second determination module is used to determine that the text content corresponding to the to-be-answered speech is not the target specific word when the target automatic speech recognition module does not output the target specific word. The target natural language understanding module includes a first mapping relationship between each preset positive word and its corresponding preset reply text (corresponding to the preset reply data of the above embodiment), a second mapping relationship between each preset negative word and its corresponding preset reply text, and the target text-to-speech module includes a third mapping relationship between each preset reply text and its corresponding reply speech.

[0115] The generation unit includes: a second input module, used to input the target specific word into the target natural language understanding module, so that the target natural language understanding module outputs a preset reply text corresponding to the target specific word according to the target specific word, the first mapping relationship and the second mapping relationship.

[0116] The third input module is used to input the preset reply text corresponding to the target specific word into the target text-to-speech module, so that the target text-to-speech module outputs the reply speech corresponding to the to-be-replied speech according to the preset reply text corresponding to the target specific word and the third mapping relationship. The first acquisition unit is used to acquire a first training sample set before the receiving unit receives the to-be-replied speech, wherein the first training sample set includes a plurality of first training samples, a plurality of second training samples, and a plurality of third training samples, the first training sample includes: any preset positive word and the speech corresponding to the preset positive word, the second training sample includes: any preset negative word and the speech corresponding to the preset negative word, and the third training sample includes: the training text that does not include any preset positive word and any preset negative word and the speech corresponding to the training text.

[0117] The first training unit is used to train the first preset model using the first training sample set until the first preset training stop condition is reached to obtain the preset acoustic model; the second acquisition unit is used to acquire the second training sample set, wherein the second training sample set includes multiple preset positive words, multiple preset negative words and multiple other words, and the other words are words related to the preset positive words or the preset negative words.

[0118] The second training unit is used to train the second preset model using the second training sample set until a second preset training stop condition is reached to obtain a preset language model.

[0119] The establishing unit is used to establish a preset recognition word network according to the inter-character relationship corresponding to each preset positive word and the inter-character relationship corresponding to each preset negative word.

[0120] In this embodiment, in step S62, problems that are difficult to solve manually are solved offline and the process ends. In the case where manual customer service is difficult to solve voice consultation, a work order is created and the work order content is processed for privacy protection. The work order is used to arrange professionals to answer the questions in the voice consultation. It should be noted that the privacy protection of the work order content can protect customer privacy, build customer trust, prevent information abuse and anonymize data.

[0121] In this embodiment, step S2 includes receiving text consultation from customers or receiving voice consultation from customers. In this way, when customers make inquiries, they can use voice consultation or text consultation (corresponding to the voice consultation content and text consultation content in the above embodiment), which is suitable for customers of all ages.

[0122] In this embodiment, in step S3, the consultation content is identified by a speech recognition model or a text recognition model. Wherein, S3 specifically includes sub-steps S31 to S36: S31: Construct an acoustic model through an input gate i, a forget gate s and an output gate h. Wherein, the input gate i is used to control the amount of information input into the acoustic model, the forget gate s is used to control the amount of information retained by the acoustic model, and the output gate h is used to control the amount of information output by the acoustic model. S32: Extract the acoustic features of the voice consultation signal frame by frame. It should be noted that the extraction of the acoustic features of the voice consultation signal frame by frame can capture the temporal nature of the voice signal, that is, the acoustic features of the voice signal will change over time. This temporal nature can provide dynamic information about the voice, such as changes in the audio spectrum over time, which helps to better understand and recognize the voice signal. S33: Input the acoustic feature vector xt into the acoustic model. Wherein, the acoustic feature vector can be one or more of an energy feature vector, a time domain feature vector, a frequency domain feature vector, a music theory feature vector and a perceptual feature vector. S34: Output the posterior probability Pt under the acoustic feature vector according to the following formula:

[0123]

[0124]

[0125]

[0126]

[0127]

[0128]

[0129]

[0130]

[0131] Among them, sigmoid () represents the curve function, r represents the information output by the acoustic model, c represents the historical memory information of the acoustic model, g represents the operation of the excitation signal, m represents the information output after passing through the forget gate, b represents the bias, α represents the weight of the input information, β represents the weight of the output information, λ represents the weight of the historical memory information, and η represents the forgetting weight. It should be noted that the formula shows that the output of the acoustic model is affected by both historical information and future information. The acoustic model can use historical information to better understand the acoustic features of the current frame. By considering the information of the previous frame, the model can capture the contextual relationship in the speech signal, thereby improving the accuracy of speech recognition. In addition, since sound signals (especially telephone voices for customer consultation) are often affected by factors such as environmental noise, microphone quality, and speaker changes, the acoustic model needs to be robust to cope with these interferences. By utilizing historical and future information, the model can better resist noise and changes, and improve the recognition accuracy and robustness of speech in complex environments. It helps to improve the accuracy of speech recognition and generate more natural and coherent speech output. S35: Convert the posterior probability P(yi|xt) into the likelihood probability P(xt|yi) through the Bayesian algorithm:

[0132]

[0133]

[0134] Among them, P(yi) represents the prior probability, which can be obtained through pre-statistical data, Tyi represents the number of frames in the data belonging to state yi, and T represents the total number of frames of data. Among them, the prior probability (PriorProbability) is a subjective estimate of the probability of an event based on past experience or previous knowledge before considering any observed data. It is the initial belief or expected probability of the possibility of an event. The prior probability does not consider the current observed data, but relies on the subjective judgment of domain experts, historical data or related information. Among them, the posterior probability (PosteriorProbability) is the conditional probability calculated based on the prior probability and current data after considering the observed data. It represents the probability of an event occurring given the observed data. By using Bayes' theorem, the posterior probability can be inferred from the prior probability and the likelihood probability. The posterior probability combines the information of the prior probability and the current data to provide an update and correction to the probability of an event occurring. Among them, the likelihood probability (LikelihoodProbability) is the possibility of the observed data occurring given a certain parameter value. The likelihood probability describes the relationship between the parameter and the data and is used to evaluate the degree of fit between a certain parameter value and the observed data. Likelihood probability is usually used for parameter estimation, and methods such as maximum likelihood estimation are used to select the parameter value that is most likely to explain the observed data. Likelihood probability does not directly represent the probability of an event, but the relative likelihood of the data occurring given the parameters. S36: Through the likelihood probability P (xt|yi) and combined with the language model, the text sequence with the highest probability is used as the recognition result of the speech recognition model through the above formula, and the recognition result is output:

[0135]

[0136] Among them, O represents the acoustic feature sequence after feature extraction, P(O) represents the probability of occurrence of a certain feature sequence, for the determined input P(O)=1, A represents the text sequence corresponding to the feature sequence, P(O|A) represents the likelihood probability, P(A) is the probability of occurrence of a certain text sequence, μ1 represents the weight of the language model in the speech recognition model, and μ2 represents the weight of the acoustic model in the speech recognition model. Among them, those skilled in the art can adjust the size of μ1 and μ2 according to actual needs, and the present invention does not limit it. In the present invention, the speech consultation is identified according to the acoustic features and language features of the speech consultation through the speech recognition model, and the text sequence with the highest probability is used as the recognition result of the speech recognition model. Compared with the single basis of acoustic features, the accuracy of speech recognition is greatly improved.

[0137] In the specific setting, in step S4, based on the preset language response model, one or more automatic reply candidate texts are generated according to the content and product information input by the customer. Among them, the language response model can be an NLP (Natural Language Generation) model. The natural language generation model may include a word vector model and an NLP task unit. The word vector model converts words in natural language (i.e., the main points of the content of the advertisement to be placed) into dense vectors. Similar words will have similar vector representations. Such conversion facilitates mining the features between words and sentences in the text. The word vector model can adopt ELMo (Embedding from Language Models), GPT (Generative Pre-Training), BERT (Bidirectional Encoder Representations from Transformers, bidirectional encoder representation from Transformer), etc. The NLP task unit converts the word vector output by the word vector model into the output of natural language, wherein the NLP task model unit can be implemented by a neural network or a linear classifier. When the word vector model is powerful enough, an output layer can be added after the word vector model as an NLP task unit.

[0138] In this embodiment, each customer consultation question is integrated again. In the process of countless repeated integrations, the intelligent reply is clearer and more intelligent, and the customer's problem is solved as quickly as possible. In the case that the robot is difficult to solve the voice consultation, the peak intelligent queuing algorithm is used to transfer to manual customer service, avoiding congestion during the consultation peak period, giving priority to the customer's urgent problems, improving the efficiency and quality of customer service, and avoiding unnecessary losses to customers.

[0139] Based on the same inventive concept, the embodiment of the present application also provides an intelligent customer service automatic recognition language reply device for implementing the intelligent customer service automatic recognition language reply method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more intelligent customer service automatic recognition language reply device embodiments provided below can refer to the limitations of the intelligent customer service automatic recognition language reply method above, and will not be repeated here.

[0140] In an exemplary embodiment, as shown in FIG5 , an intelligent customer service automatic language recognition reply device 500 is provided, comprising: a consultation receiving module 502 , a feature recognition module 504 and a consultation reply module 506 , wherein:

[0141] The consultation receiving module 502 is used to receive consultation content initiated by the user through the user terminal; the consultation content includes voice consultation content and / or text consultation content.

[0142] The feature recognition module 504 is used to recognize the consultation content using a language model and / or an acoustic model to obtain language features corresponding to the consultation content; the language features represent the consultation intention corresponding to the consultation content.

[0143] The consultation reply module 506 is used to retrieve the preset reply data based on the language features using the neighbor interpolation algorithm, and when the target preset reply is retrieved, the target preset reply is sent to the user terminal; the preset reply data is obtained based on the integration of historical consultation content and historical reply data.

[0144] Furthermore, in one embodiment, the feature recognition module 504 is also used to, when the consultation content includes voice consultation content, use an acoustic model to process the consultation content to obtain consultation voice data; use a language model to recognize the consultation text data to obtain language features; and when the consultation content includes text consultation content, use a language model to recognize the text consultation content to obtain language features.

[0145] Furthermore, in one embodiment, the consultation response module 506 is also used to determine the characteristic words included in the language features; when the characteristic words are affirmative words, a neighbor interpolation algorithm is used to retrieve the target affirmative characteristic words from a preset characteristic word library; based on the target affirmative characteristic words and the first mapping relationship, the preset response data is retrieved to obtain the target preset response.

[0146] Furthermore, in one embodiment, the consultation reply module 506 is also used to retrieve a target negative characteristic word from a preset characteristic word library using a neighbor interpolation algorithm when the characteristic word is a negative word; and retrieve the preset reply data based on the target negative characteristic word and the second mapping relationship to obtain a target preset reply.

[0147] Furthermore, in one embodiment, the intelligent customer service automatic recognition language reply device 500 also includes: a mapping construction module, which is used to construct a first mapping relationship based on the positive characteristic words included in the preset characteristic vocabulary and the preset reply data corresponding to the positive characteristic words; construct a second mapping relationship based on the negative characteristic words included in the preset characteristic vocabulary and the preset reply data corresponding to the negative characteristic words; construct a third mapping relationship based on all preset reply data and the voice replies corresponding to the preset reply data; the third mapping relationship is used to send the voice reply corresponding to the target preset reply to the user terminal when the consultation content contains voice consultation content.

[0148] Furthermore, in one embodiment, the consultation reply module 506 is also used to call the manual customer service interface according to the peak period intelligent queuing algorithm to connect the communication between the user terminal and the manual customer service when the target preset reply is not retrieved; obtain the reply data of the manual customer service after the communication is connected through the manual customer service interface, associate the reply data of the manual customer service interface with the language features, and update the preset reply data.

[0149] Each module in the above-mentioned intelligent customer service automatic recognition language reply device 500 can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0150] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG6. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as consultation content, language features, and preset reply data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for automatically identifying language replies by intelligent customer service is implemented.

[0151] Those skilled in the art will understand that the structure shown in FIG. 6 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0152] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0153] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0154] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0155] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0156] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0157] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0158] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for automatically identifying language responses by intelligent customer service, characterized in that: The method comprises: Receiving consultation content initiated by a user through a user terminal; the consultation content includes voice consultation content and / or text consultation content; The consultation content is identified by using a language model and / or an acoustic model to obtain language features corresponding to the consultation content; the language features represent the consultation intention corresponding to the consultation content; According to the language features, a neighbor interpolation algorithm is used to retrieve the preset reply data. When the target preset reply is retrieved, the target preset reply is sent to the user terminal; the preset reply data is obtained based on the integration of historical consultation content and historical reply data.

2. The method according to claim 1, characterized in that The adopting of a language model or an acoustic model to identify the consultation content to obtain language features corresponding to the consultation content includes: In the case where the consultation content includes voice consultation content, the consultation content is processed using an acoustic model to obtain consultation voice data; Using a language model to recognize the consultation voice data to obtain language features; In the case where the consultation content includes text consultation content, a language model is used to identify the text consultation content to obtain language features.

3. The method according to claim 1, characterized in that The step of retrieving the preset reply data using a neighbor interpolation algorithm according to the language feature includes: Determining the characteristic words included in the language feature; In the case where the characteristic word is a positive characteristic word, a neighbor interpolation algorithm is used to retrieve a target positive characteristic word from a preset characteristic word library; According to the target positive characteristic word and the first mapping relationship, the preset reply data is retrieved to obtain the target preset reply.

4. The method according to claim 3, characterized in that The method further comprises: In the case where the characteristic word is a negative characteristic word, a neighbor interpolation algorithm is used to retrieve a target negative characteristic word from a preset characteristic word library; According to the target negative characteristic word and the second mapping relationship, the preset reply data is retrieved to obtain the target preset reply.

5. The method according to claim 4, characterized in that Before searching the preset reply data using the neighbor interpolation algorithm according to the language feature, the method further includes: Constructing a first mapping relationship according to the positive characteristic words included in the preset characteristic word library and the preset reply data corresponding to the positive characteristic words; Constructing a second mapping relationship according to the negative characteristic words included in the preset characteristic word library and the preset reply data corresponding to the negative characteristic words; A third mapping relationship is constructed based on all the preset reply data and the voice replies corresponding to the preset reply data; the third mapping relationship is used to send the voice reply corresponding to the target preset reply to the user terminal when the consultation content includes voice consultation content.

6. The method according to claim 1, characterized in that The method further comprises: In the case where the target preset reply is not retrieved, calling the manual customer service interface according to the peak intelligent queuing algorithm to connect the communication between the user terminal and the manual customer service; The reply data of the manual customer service after the communication is connected is obtained through the manual customer service interface, the reply data of the manual customer service interface is associated with the language feature, and the preset reply data is updated.

7. An intelligent customer service automatic recognition language reply device, characterized in that: The device comprises: A consultation receiving module, used to receive consultation content initiated by a user through a user terminal; the consultation content includes voice consultation content and / or text consultation content; A feature recognition module is used to recognize the consulting content by using a language model and / or an acoustic model to obtain language features corresponding to the consulting content; the language features represent the consulting intention corresponding to the consulting content; The consultation reply module is used to retrieve the preset reply data according to the language features using a neighbor interpolation algorithm, and when the target preset reply is retrieved, the target preset reply is sent to the user terminal; the preset reply data is obtained based on the integration of historical consultation content and historical reply data.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

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

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images