Auxiliary method and device for online customer service reply and storage medium

Abstract

The invention discloses an auxiliary method for online customer service reply, relates to the technical field of intelligent equipment control, and discloses an auxiliary method for online customer service reply, comprising: acquiring real-time inquiry information of a current work order; filtering the real-time inquiry information by using a preset rule engine to obtain inquiry content corresponding to the user service demand; the inquiry content is input into a big data model, knowledge points corresponding to the inquiry content output by the big data model based on a preset knowledge base are obtained, and the knowledge points comprise content meeting user service requirements. The invention further discloses an auxiliary device for online customer service reply and a storage medium, a natural language processing technology, big data analysis and an artificial intelligence algorithm are introduced, rapid and accurate answering of customer questions is achieved, and meanwhile through continuous optimization and learning, the working efficiency and service quality of customer service staff are improved.

Description

Technical Field

[0001] The present application relates to the field of big data applications, and in particular to an auxiliary method, device and storage medium for online customer service replies. Background Art

[0002] In traditional customer service work, faced with a large number of repetitive questions raised by customers, customer service staff often need to answer them one by one, which not only consumes a lot of time and energy, resulting in low work efficiency, but also causes fatigue to customer service staff due to long-term repetitive work. In addition, due to differences in the experience, professional knowledge and coping ability of the agents, the quality of the answers they provide varies, which to a certain extent affects the overall service experience of customers and may even lead to customer dissatisfaction and complaints due to manual errors.

[0003] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0004] The main purpose of this application is to provide an auxiliary method, device and storage medium for online customer service replies, aiming to solve the technical problem that the quality of answers that existing online customer service can provide when receiving user inquiries is uneven, affecting the customer's overall service experience.

[0005] To achieve the above purpose, the present application proposes an auxiliary method for online customer service reply, and the auxiliary method for online customer service reply includes:

[0006] Get real-time inquiry information of the current work order;

[0007] Filter the real-time inquiry information with a preset rule engine to obtain inquiry content corresponding to the user's business needs;

[0008] The query content is input into a big data model, and knowledge points corresponding to the query content output by the big data model based on a preset knowledge base are obtained, where the knowledge points include content that solves user business needs.

[0009] In one embodiment, after the step of obtaining the knowledge points corresponding to the query content output by the big data model based on the preset knowledge base, the step further includes:

[0010] In response to the editing operation on the knowledge point, answer information corresponding to the inquiry information is determined according to the edited knowledge point and outputted.

[0011] In one embodiment, the step of inputting the query content into a big data model and obtaining knowledge points corresponding to the query content output by the big data model based on a preset knowledge base includes:

[0012] Processing the query content with the multimodal feature extraction and fusion algorithm of the big data model to obtain the user's business needs;

[0013] The knowledge points are acquired from the preset knowledge base based on the user's business requirements.

[0014] In one embodiment, the step of processing the query content with the multimodal feature extraction and fusion algorithm of the big data model to obtain the user's business needs includes:

[0015] Obtaining characteristic information according to the processing result of the inquiry content;

[0016] The feature information is input into the semantic network of the big data model, and the user business needs represented by the feature information are determined by the multi-level association rules of the semantic network.

[0017] In one embodiment, the auxiliary method for online customer service reply further includes:

[0018] Determine the user's business requirement at the inquiry location of the current work order;

[0019] The knowledge point is marked at the inquiry position, and an association relationship between the work order and the knowledge point is created and saved.

[0020] In one embodiment, the auxiliary method for online customer service reply further includes:

[0021] When the knowledge points corresponding to the query content output by the big data model based on the preset knowledge base are not obtained, determining the user business needs of the query content;

[0022] Create content that solves the user's business needs and generate new knowledge point output based on the content.

[0023] In one embodiment, after the step of creating content that solves the user's business needs and generating knowledge point output based on the content, the step further includes:

[0024] Determining whether the preset knowledge base stores a target knowledge point that is repeated with the new knowledge point;

[0025] When it is determined that the target knowledge point is not stored in the preset knowledge base, storing the new knowledge point in the preset knowledge base;

[0026] When it is determined that the target knowledge point is stored in the preset knowledge base, the user business demand corresponding to the new knowledge point is added to the storage content corresponding to the target knowledge point.

[0027] In one embodiment, the auxiliary method for online customer service reply further includes:

[0028] Creating an initial data model, and training the initial data model using training samples generated by historical work orders;

[0029] The large data model is obtained based on the training result of the initial data model.

[0030] In addition, to achieve the above-mentioned purpose, the present application also proposes an auxiliary device for online customer service response, the device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the auxiliary method for online customer service response as described above.

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

[0032] One or more technical solutions proposed in this application have at least the following technical effects:

[0033] The technical solution of this application obtains the real-time inquiry information of the current work order; filters the real-time inquiry information with a preset rule engine to obtain the inquiry content corresponding to the user's business needs; inputs the inquiry content into a big data model to obtain the knowledge points corresponding to the inquiry content output by the big data model based on a preset knowledge base, and the knowledge points include content that solves the user's business needs. By introducing natural language processing technology, big data analysis and artificial intelligence algorithms, fast and accurate answers to customer questions are achieved, and through continuous optimization and learning, the work efficiency and service quality of customer service personnel are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0035] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0036] Figure 1 A flowchart of the first embodiment of the auxiliary method for online customer service reply of this application is provided;

[0037] Figure 2 A flowchart diagram of Embodiment 2 of the auxiliary method for online customer service reply of this application;

[0038] Figure 3 A flowchart diagram of Embodiment 3 of the auxiliary method for online customer service reply of this application;

[0039] Figure 4 A schematic diagram of the device structure of the hardware operating environment involved in the auxiliary method for online customer service reply in an embodiment of the present application.

[0040] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0041] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0042] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0043] The main solution of the embodiment of the present application is: obtaining real-time inquiry information of the current work order; filtering the real-time inquiry information with a preset rule engine to obtain inquiry content corresponding to the user's business needs; inputting the inquiry content into a big data model, and obtaining knowledge points corresponding to the inquiry content output by the big data model based on a preset knowledge base, wherein the knowledge points include content that solves the user's business needs.

[0044] Due to differences in the experience, expertise and response capabilities of current agents, the quality of answers they can provide when receiving user inquiries varies, which to some extent affects the resolution of customer business needs and thus reduces the user's overall service experience.

[0045] This application provides a solution that achieves fast and accurate answers to customer questions by introducing natural language processing technology, big data analysis and artificial intelligence algorithms. At the same time, through continuous optimization and learning, it improves the work efficiency and service quality of customer service personnel.

[0046] Based on this, the embodiment of the present application provides an auxiliary method for online customer service reply, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the auxiliary method for online customer service reply in this application.

[0047] In this embodiment, the auxiliary method for online customer service reply includes steps S10 to S30:

[0048] Step S10, obtaining real-time inquiry information of the current work order;

[0049] Step S20, filtering the real-time inquiry information with a preset rule engine to obtain inquiry content corresponding to the user's business needs;

[0050] Step S30, inputting the query content into a big data model, obtaining knowledge points corresponding to the query content output by the big data model based on a preset knowledge base, wherein the knowledge points include content that solves user business needs.

[0051] In this embodiment, when accepting the user's inquiry, the online customer service creates a work order between the online customer service and the user, and saves the work order when the user disconnects the inquiry connection with the online customer service. Before creating the work order, the user triggers the contact window opening instruction by entering the relevant function page, thereby opening the contact window for contacting the online customer service, and initiating an inquiry based on the contact window.

[0052] In a further embodiment, an enterprise or a business deploys a relevant system platform and opens a contact window, so that the user can open the contact window through the relevant function button and input relevant inquiry information for consultation. In addition, based on the deployed contact window, an automatic reply item is provided. After the contact window detects that the user has initiated an inquiry, the set automatic reply item is automatically output. The automatic reply item is in the form of a link, and the user can obtain the actual solution of the automatic reply item by triggering a certain automatic reply item. In actual application, the automatic reply items are respectively represented as related business needs. Therefore, the actual solution output based on triggering the automatic reply item is also to solve the business needs represented by the automatic reply item.

[0053] In this embodiment, a knowledge base is provided, and the substantial solution based on the triggering output of the automatic reply item is a knowledge point pre-stored in the preset knowledge base, and the knowledge point, when stored, includes product information, services, common problems and solutions, etc. Therefore, the process of outputting the substantial solution based on the automatic reply item is essentially to analyze the business needs, and select the knowledge points matching the business needs from the knowledge base according to the analysis results, and output the solutions possessed by the selected knowledge points.

[0054] In addition, in order to improve the reply speed of the automatic reply item, the solution of the automatic reply item can be stored in advance as cache data for quick output. That is, a cache data storage area is provided for storing the solution of the automatic reply item, or for storing the knowledge points corresponding to the automatic reply item. The user can update the cache data storage area based on the actual situation, thereby updating the business needs represented by the automatic reply item.

[0055] After the user obtains a solution to the current business needs through the automatic reply item in the contact window, the user can directly exit the contact window; however, after the user cannot obtain a solution to the current business needs based on the automatic reply item, the user can directly connect with the online customer service for a manual reply. Alternatively, after the user stays in the contact window for a preset time, the user can connect with the online customer service through the contact window to provide a solution to the current business needs of the user through the online customer service.

[0056] To this end, when the user successfully connects with the online customer service through the current contact window, a work order based on the online customer service and the user is created. The work order is essentially a record of a single consultation operation between the customer service and the user. The specific record information includes customer problems, solutions, etc. Therefore, the work order can also be defined as a consulting work order.

[0057] The work order is automatically generated after the user and the online customer service are connected in communication, and the work order is stored after the user and the online customer service are disconnected from the communication connection. In actual applications, user consultation can be characterized by communication forms such as client contact windows, web pages, and voice calls. Therefore, based on the differences in the communication forms, the work order can be created when the user is connected to the online customer service for communication, and the work order can be saved when the communication is disconnected. For example, when the user connects to the online customer service through the contact window, a work order for the current inquiry between the user and the online customer service is created. When the user closes the contact window, it is determined that the current communication connection is disconnected, and the work order is saved. In further inquiry scenarios, the creation and saving actions of the work order can be determined based on the communication connection method represented by the current inquiry scenario.

[0058] As shown above, based on the communication connection between the user and the online customer service, a work order for this inquiry is created, so as to collect the inquiry information between the user and the online customer service through the creation of the work order. Since the work order is created after the current user connects with the online customer service, the inquiry information collected based on the work order should be real-time inquiry information.

[0059] In this embodiment, the acquired real-time inquiry information collected by the current work order is filtered to obtain the user's inquiry content based on the real-time inquiry information. This is because when the user consults with the online customer service, the solution based on the automatic reply item may fail to solve the user's current business needs, or the user may not be able to clearly define the business needs, so he or she will connect with the online customer service. Regardless of the reason for initiating the connection with the online customer service, the real-time inquiry information generated is the conversation with the online customer service. The actual conversation contains language that cannot represent business needs. In order to reduce the data processing volume of the big data model, the real-time inquiry information is filtered by a preset rule engine to obtain inquiry content that may contain business needs.

[0060] The setting of the preset rule engine is essentially a data processing mechanism generated by defining rules, which is used to filter the conversation content in the dialogue system to improve the response quality and efficiency of the system. The essential setting steps of the preset rule engine are as follows:

[0061] 1) Define rules: Used to define rules to identify different types of conversations. The rules can be created based on keywords, phrases, regular expressions, or natural language processing (NLP) technology. For example, Rule 1 classifies a conversation as "order issue" if it contains keywords such as "refund", "return", or "cancel order".

[0062] 2) Configure the rule engine: used to input the defined rules into the rule engine and specify the triggering conditions and corresponding actions (such as classification, reply or transfer) of each rule.

[0063] 3) Integrated dialogue system: used to integrate the rule engine with the dialogue system (such as a chatbot). The integration can be achieved through AP calls, message queues, or middleware to ensure that after receiving user input, the dialogue system can pass it to the rule engine for processing.

[0064] In the above, based on the preset rule engine, the conversation is processed based on the acquired real-time inquiry information. In this embodiment, it is limited to that when the user interacts with the online customer service, the real-time inquiry information is extracted and sent to the preset rule engine. The preset rule engine classifies the real-time inquiry information according to the set rules and triggers corresponding actions. For example, if the conversation is classified as "order problem", the chatbot can automatically provide relevant help for refund, return or cancellation of order; if the conversation is classified as "other problems", the chatbot can transfer the conversation to manual customer service for processing.

[0065] Furthermore, based on the application of the preset rule engine, it is also necessary to regularly monitor the performance of the preset rule engine and the feedback of the dialogue system. The feedback of the dialogue system is generally reported through manual customer service, or a functional questionnaire is set for manual customer service to fill in in order to collect feedback on the application of the preset rule engine. Based on this, it is possible to verify whether the rules set in the preset search engine do not conform to the actual application situation according to the feedback, for example, it is found that some rules are inaccurate or too complicated. In this way, the preset rule engine is adjusted and optimized according to the feedback. Alternatively, new rules are added according to the feedback or the filtering data analysis of the preset rule engine.

[0066] In the application of the preset rule engine, it is necessary to ensure that the appropriate priority is set for the set rules so that the optimal processing action can be selected when multiple rules are matched. In addition, avoid creating overly complex rules, which will lead to performance degradation and difficulty in maintenance. Furthermore, when filtering the real-time inquiry information, it is also necessary to ensure compliance with relevant laws and regulations and user privacy policies.

[0067] In the above, after filtering the real-time query information through the preset rule engine, based on the filtering result of the real-time query information, the filtering result is defined as the query content containing the user's business needs, that is, the query content is input into the preset big data model for processing, so as to obtain the user's business needs based on the current query operation, that is, the user's business needs. The data structure in the preset big data model can process the query content with the user's business needs to obtain the user's business needs.

[0068] The preset big data model has a language recognition algorithm, which recognizes the filtered query content and determines the user's business needs based on the current query operation according to the recognition result. The language recognition algorithm can also be used as a content recognition process for the query content.

[0069] In a feasible implementation, step S20 may include steps A21 to A22: the step of inputting the query content into a big data model and obtaining knowledge points corresponding to the query content output by the big data model based on a preset knowledge base includes:

[0070] A21. Process the query content using the multimodal feature extraction and fusion algorithm of the big data model to obtain the user's business needs;

[0071] A22. Acquire knowledge points from the preset knowledge base based on the user's business needs.

[0072] In this embodiment, a multimodal feature extraction and fusion algorithm is deployed in the big data model, and the query content is processed by the multimodal feature extraction and fusion algorithm to obtain the business requirements contained therein.

[0073] The multimodal feature extraction and fusion algorithm deployed in the big data model includes a multimodal feature extractor and a multimodal feature fuser, wherein the multimodal feature extractor can extract useful feature representations from original data of different modalities, and the multimodal feature fuser is used to fuse the feature representations extracted by the multimodal feature extractor into a unified feature vector, so as to determine the actual business needs with the feature vector.

[0074] In practical applications, the multimodal feature extractor can extract features of different data types based on the information representation form of the query content, and the existing representation forms that can realize query interaction may include voice information, text information, and image features. Therefore, when the multimodal feature extractor extracts features from the query content, it needs to process them separately based on the representation form of the query content, that is, voice feature extraction is performed on the query content in voice format, text feature extraction is performed on the query content in text format, and image feature extraction is performed on the query content in graphic format. The specific feature extraction steps can be as follows:

[0075] 1) Extracting speech features based on the inquiry content in voice format, and processing the speech signal of the inquiry content using preprocessing techniques (such as denoising, framing, windowing, etc.); extracting acoustic features, such as Mel-frequency cepstral coefficients (MFCC), spectrum envelope, fundamental frequency, etc., and using deep learning models (such as convolutional neural networks (CNN) or recurrent neural networks (RNN)) to automatically extract high-level speech features;

[0076] 2) Extract text features based on the query content in text format, perform preprocessing on the text such as word segmentation, stop word removal, and stem extraction; use methods such as bag-of-words model, TF-IDF, word embedding (such as Word2Vec, BERT) to extract text features; consider the semantic information and syntactic structure of the text, and use natural language processing technology (such as syntactic analysis, semantic role labeling) to further enrich the features;

[0077] 3) Image feature extraction based on the query content in graphic format requires preprocessing of the image, such as scaling, cropping, denoising, etc., and the use of traditional image processing techniques (such as edge detection, texture analysis) to extract low-level features; deep learning models (such as convolutional neural networks (CNN)) are used to extract high-level features, such as local features (such as SIFT, SURF) and global features (such as global average pooling features).

[0078] In the above, the query content is subjected to feature extraction based on different data types according to the multimodal feature extractor. After obtaining the feature information of the query content, the extracted feature information is fused by the multimodal feature fusion device. During the fusion process, since the feature vectors of different modalities may have different dimensions and representations, feature alignment is required. A dimensionality reduction technique (such as principal component analysis PCA and linear discriminant analysis LDA) is used to map the feature vectors to the same low-dimensional space; alternatively, a deep learning model (such as an encoder) is used to learn the mapping relationship between features of different modalities to perform feature alignment.

[0079] And, the feature fusion strategy based on the feature information of the query content includes but is not limited to splicing fusion, weighted fusion, feature conversion, graph fusion and unified feature vector. In the specific implementation process, the actual operation based on the feature fusion strategy is: splicing fusion, directly splicing the feature information of different modalities to form a longer feature information, wherein the feature information can also be represented as a feature vector, that is, multiple feature vectors of different modalities are spliced ​​together to form a longer feature vector; weighted fusion is to assign different weights to different feature information according to the importance or reliability of each modality; feature conversion can use deep learning models (such as multi-layer perceptron MLP, attention mechanism) to learn the interaction and fusion mode between different modal features; graph fusion is based on the existence of a clear structural relationship between each feature information, such as the object correspondence between text and image, and uses graph neural network (GNN) for feature fusion; unified feature vector is to obtain unified feature information after feature fusion of multiple feature information, that is, the feature information contains multiple feature information from different modalities.

[0080] Furthermore, the fused feature information may be further processed, such as normalization, standardization and other processing benchmarks, so that the processed feature information has more complete consistency and comparability in the subsequent recognition and matching process, thereby improving the recognition accuracy of user business needs.

[0081] In addition, in a feasible implementation, step A22 may further include steps A22-1 and A22-2, that is, the step of processing the query content with the multimodal feature extraction and fusion algorithm of the big data model to obtain the user's business needs includes:

[0082] A22-1. Obtaining characteristic information according to the processing result of the inquiry content;

[0083] A22-2. Input the feature information into the semantic network of the big data model, and determine the user business needs represented by the feature information using the multi-level association rules of the semantic network.

[0084] In this embodiment, after processing the query content based on the multimodal feature extraction and fusion algorithm and obtaining the feature information of the query content based on the processing result, in the process of determining the business needs based on the feature information, since the feature information is extracted from the query content, the query information, after filtering based on the real-time query information, results in no semantic connection between the query content, or may have multiple words with or without connection. Therefore, the feature information can be associated based on the semantic network set in the big data model, thereby forming the user business needs represented by the feature information.

[0085] It should be clarified that the semantic network is a language network system with multi-level association rules, and is deployed in the big data model to determine business needs. Based on the definition of the semantic network, the semantic network with multi-level association rules is a composite data structure that combines multi-level association rules with the concept of semantic network. The multi-level association rules are an extension of the analysis based on traditional association rules, and their purpose is to find each associated item set at multiple levels. In this application, they are used to find item sets associated with current feature information at multiple levels. The multi-level association rules are different from traditional association rule analysis, and the difference lies mainly in that the traditional association rule analysis mainly focuses on the association relationship between items, while the multi-level association rules also take into account the hierarchical structure of items.

[0086] Therefore, it can be known that the application of the multi-level association rules can be used for the relationship analysis of multiple feature information, the semantic level division is performed based on the multiple feature information, and the relevant level analysis is performed according to the semantic level of each feature information, so as to obtain the cross-level association rules of the feature information, and then the corresponding feature information is associated based on the association rules to obtain the user business needs that can be represented by the multiple feature information. This process can deeply capture the hidden information possessed by each feature information, so as to more accurately determine the user business needs through the feature information, that is, improve the accuracy of user business needs confirmation.

[0087] As shown above, the multi-level association rule is essentially a data rule, which needs to be deployed in an actually available data structure for application, that is, the multi-level association rule is set in the semantic network deployed by the big data model for application. The semantic network is a semantic-oriented structure, which is essentially a process of using a set of inference rules to process special arcs in the network. In practical applications, the semantic network can provide semantic relationships with concepts for typical declarative sentences, thereby organizing multiple words into information that simulates human semantic memory and association methods.

[0088] That is, after the query content is processed by a multimodal feature extraction and fusion algorithm to obtain feature information, it is input into a semantic network, and the information is combined through the multi-level association rules of the semantic network to obtain accurate user business needs. The user business needs can be represented by multiple words, or text paragraphs, etc. that can be used to represent business needs, and the business needs included in the user business needs are one or more.

[0089] The above are only two feasible implementations of step S20 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S20.

[0090] In summary, after determining the user's business needs by filtering the obtained real-time inquiry information based on the preset big data model, the knowledge points corresponding to the business needs can be obtained from the knowledge base based on the user's business needs as a condition, and the knowledge points can be output. When outputting the knowledge points, they can be output based on the communication connection between the online customer service represented by the work order and the user. For example, the online customer service can output the obtained knowledge points through the contact window opened between the online customer service and the user.

[0091] In addition, after obtaining the knowledge points, the online customer can also edit and output the knowledge points to make them more in line with the user's business needs and the current conversation scenario, that is, after the step of obtaining the knowledge points corresponding to the call content output by the big data model based on the preset knowledge base, it also includes:

[0092] In response to the editing operation on the knowledge point, answer information corresponding to the inquiry information is determined according to the edited knowledge point and outputted.

[0093] After processing the query content through the big data model, the user's business needs based on the query content are obtained, and the corresponding knowledge points are screened in the preset knowledge base based on the user's business needs. The knowledge points are characterized as data information with user business solutions, including the indicated business needs, product information, services, common problems and solutions. To this end, the screening process is essentially a keyword matching process based on the user's business needs, that is, the knowledge points in the preset knowledge base that contain the user's business needs keywords are output as screening results.

[0094] Therefore, based on the output of the screened knowledge points, the manual customer service can edit the output knowledge points to edit the content of the knowledge points into reply information output that better meets the user's business needs and the current dialogue scenario, thereby improving the user's business needs and further improving the user experience.

[0095] Further, in another feasible implementation of the present application, in order to integrate the information of user needs and make relevant statistics on user needs, the knowledge point needs to be associated with the work order and saved, that is, the auxiliary method of online customer service reply, further includes:

[0096] Determine the user's business requirement at the inquiry location of the current work order;

[0097] The knowledge point is marked at the inquiry position, and an association relationship between the work order and the knowledge point is created and saved.

[0098] Since the real-time inquiry information is the communication information formed after the user connects with the online customer service, therefore, when the user's business needs are obtained through processing based on the real-time inquiry information, and the corresponding knowledge points are obtained from the knowledge base based on the user's business needs, it is necessary to add a knowledge point mark to the real-time inquiry information location where the inquiry content corresponding to the user's needs is located, so as to retain the conversation proof of the user's business needs. That is, it can be obtained that the location of the inquiry content corresponding to the user's business needs in the real-time inquiry information is added, and a knowledge point mark is added. In other words, when the user's business needs are determined based on the currently identified real-time inquiry information, a knowledge point mark is added at the real-time inquiry information.

[0099] The newly added knowledge point mark is a symbol or annotation with marking capability, which can be used to represent the current real-time query information to obtain user business needs and obtain knowledge points. To this end, the unique identification information of the knowledge point can be used as the knowledge point mark, such as the knowledge point number.

[0100] In addition, the real-time inquiry information is generated after the current user connects with the online customer service, so continuous conversation interaction will occur, thereby generating continuous real-time inquiry information. Therefore, there may be multiple user business needs obtained by filtering and analyzing the real-time inquiry information, and there are also multiple knowledge points obtained based on the business needs, so there are also multiple knowledge point tags added in the real-time inquiry information. In order to improve the directionality of the business needs, when the work order is saved, the knowledge points are determined according to the knowledge point tags, and the work order and the knowledge points are associated and saved, so that the corresponding knowledge points can be quickly determined based on the work order, so as to quickly count the knowledge points and the business needs corresponding to the knowledge points.

[0101] In this embodiment, by introducing natural language processing technology, big data analysis and artificial intelligence algorithms, fast and accurate answers to customer questions are achieved. At the same time, through continuous optimization and learning, the work efficiency and service quality of customer service personnel are improved.

[0102] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated in the following. Figure 2 , Figure 2 A flowchart of Embodiment 2 of the auxiliary method for online customer service reply of this application is provided. The auxiliary method for online customer service reply also includes:

[0103] Step S40: when the knowledge point corresponding to the query content output by the big data model based on the preset knowledge base is not obtained, determining the user business demand of the query content;

[0104] Step S50: Create content that solves the user's business needs, and generate knowledge point output based on the content.

[0105] Step S60, determining whether the preset knowledge base stores a target knowledge point that is repeated with the new knowledge point;

[0106] Step S70: When it is determined that the target knowledge point is not stored in the preset knowledge base, storing the new knowledge point in the preset knowledge base;

[0107] Step S80: When it is determined that the target knowledge point is stored in the preset knowledge base, the user business demand corresponding to the new knowledge point is added to the storage content corresponding to the target knowledge point.

[0108] In this embodiment, after filtering the real-time query information through the preset rule engine and obtaining the user business needs after processing the filtered query content through the big data model, when obtaining the corresponding knowledge points from the knowledge base based on the user business needs, there may be a situation where no knowledge points are obtained. Based on this, relevant content that can solve the user business needs is recreated as a solution for the user business needs, and new knowledge point output is generated with the newly created solution.

[0109] Among them, the solution recreated based on the user's business needs can be created based on the work experience and ability of the online customer service, or the solution can be generated by obtaining corresponding knowledge points according to historical related business needs and making changes. Its specific operation can be determined based on the solution steps or solution process of the relevant business needs. In this way, after obtaining the solution corresponding to the business needs, the solution generates new knowledge points for output.

[0110] In order to increase the number of knowledge points stored in the preset knowledge base, after the new knowledge points are created, the new knowledge points are stored in the knowledge base, thereby increasing the number of knowledge points stored in the preset knowledge base, thereby expanding the scope of business needs that can be solved. Wherein, based on the storage operation of the preset knowledge base, when generating knowledge points from the created content that can solve the user's business needs, it is necessary to generate them based on the generation format of the knowledge points. Wherein, the format of the new knowledge point is related to the data storage rules of the preset knowledge base, that is, the solution obtained is generated into a new knowledge point through the data storage rules of the preset knowledge base.

[0111] In this implementation, the preset knowledge base is presented in the form of a database, such as MySQL, PostgreSQL, Oracle, SQL Server, etc. Based on the data storage rules set in the form, the relevant information contained in the knowledge point includes the indicated business needs, product information, services, common problems and solutions, and the divided information can be stored in the form and associated respectively. The specific form association relationship is related to the database form to which the preset knowledge base belongs, and will not be elaborated here. Therefore, when the generated new knowledge point is stored in the knowledge base, it is actually a data update operation of the preset knowledge base.

[0112] In addition, the possibility that the user's business needs fail to obtain corresponding knowledge points is that the preset knowledge base fails to store corresponding knowledge points, or the content of the business needs indicated by the user in the relevant knowledge points fails to register the business needs, or the characterization information based on the business needs is different from the business needs registered by the corresponding knowledge points, resulting in the failure to obtain the corresponding knowledge points. Therefore, when storing the new knowledge point in the preset knowledge base, it is also necessary to determine whether there are knowledge points in the preset knowledge base that are repeated with the new knowledge point. As a distinction, in this embodiment, the knowledge points in the preset knowledge base that are similar to the new knowledge point are defined as target knowledge points.

[0113] Furthermore, based on the determination that the target knowledge point is similar to the new knowledge point, the new knowledge point can be matched with the target knowledge point, and the similarity data of the two can be obtained according to the matching result, and when the similarity data reaches a preset value, the two are determined to be similar, wherein the matching information focuses on the similarity of the solutions possessed by the two. In this way, in order to avoid the content stored in the preset knowledge base from repeatedly occupying storage space, and to avoid misleading online customer service due to duplicate knowledge points, the business requirements corresponding to the new knowledge point are updated to the content indicating the business requirements included in the target knowledge point, thereby avoiding duplicate storage of knowledge points. Alternatively, the new knowledge point and the target knowledge point are screened for duplicate content, so that the content that is not duplicated between the new knowledge point and the target knowledge point is added to the target knowledge point, so as to update and upgrade the content represented by the target knowledge point.

[0114] Further, since the new knowledge point is generated based on the corresponding business demand, the real-time inquiry information where the business demand is located should be marked with the new knowledge point, or the new knowledge point has not been marked. After combining the new knowledge point with the target knowledge point, the target knowledge point is used as the corresponding knowledge point of the user's business demand, that is, the target knowledge point is marked at the real-time inquiry information position where the inquiry content indicated by the user's business demand is located, and when the work order is saved, a new association relationship between the work order and the target knowledge point is created. Further, the marking information of the new knowledge point is traversed, and the markings of the new knowledge point are all changed to the target knowledge point to accurately locate the content of the knowledge point. In this embodiment, when the knowledge point is not obtained, a new solution is created based on the business demand, and the solution generates a new knowledge point, thereby increasing the scope of the business demand that can be solved. In addition, when the new knowledge point is updated to the knowledge base, the waste of storage area caused by the storage of repeated knowledge points can be avoided to the greatest extent, and further, the business demand that can be processed by the knowledge point can be accurately located, thereby improving the solution effect of the business demand.

[0115] For further information, please refer to Figure 3 , Figure 3 The flowchart of Embodiment 3 of the auxiliary method for online customer service reply of this application is provided. The auxiliary method for online customer service reply also includes:

[0116] Step S90: Create an initial data model, and train the initial data model using training samples generated by historical work orders;

[0117] Step S100: obtaining the big data model based on the training result of the initial data model.

[0118] In this embodiment, the big data model is obtained after training the data model based on the data structure in the preset initial data model by generating training samples with historical work orders. Among them, when the big data model is used to filter real-time inquiry information, its main filtering ability is to eliminate inquiry information that is not related to business needs, thereby realizing the screening of inquiry information that may have business needs. This filtering ability involves language recognition and semantic recognition. Therefore, the data structure of the initial data model has a multimodal feature extraction and fusion algorithm and a semantic network, and the specific application is realized by the multimodal feature extractor and multimodal feature fuser of the multimodal feature extraction and fusion algorithm. In addition, semantic mining is performed based on the multi-level association rules provided in the semantic network, so as to extract relevant content with business needs from the real-time inquiry information shown.

[0119] Therefore, the process of training the initial data model is essentially to train the data structure composed of the modal feature extraction fusion algorithm and the semantic network in the initial data model. That is, the pre-stored historical work order information is obtained to generate corresponding training samples to train the initial data model to obtain the big data model.

[0120] In this embodiment, training based on a big data model is a complex and resource-intensive process involving multiple key steps and professional techniques, as shown below:

[0121] 1) Data preparation for training samples

[0122] Data collection: A large number of historical work orders are collected from various sources, wherein the historical work orders contain a large amount of structured data (such as a relational database for storing data associations with knowledge points) and unstructured data (such as text, images, videos, etc.).

[0123] Data cleaning: Clean and preprocess the acquired historical work orders, including deleting duplicate data, filling missing values, processing outliers, etc., to ensure the quality and applicability of the data.

[0124] Data conversion and expansion: The original data of the historical work orders is converted into features that can be used by the machine learning algorithm. The specific features can be derived based on the information that the machine learning algorithm can recognize, such as extracting word bags in text, pixel values ​​in images, etc. Furthermore, data enhancement technology can be used to expand the training data set formed by the historical work orders to improve the generalization ability of the initial data model.

[0125] 2) Model design

[0126] Select model architecture: Design a suitable neural network model structure according to task requirements, and set a multimodal feature extraction fusion algorithm and a semantic network in the neural network model structure. In addition, based on the data volume of the initial data model, a deeper network structure, residual connection, batch normalization and other technologies can also be used to improve the expression ability and training efficiency of the initial data model.

[0127] Initialization parameters: Before training the initial data model, the parameters of the initial data model are initialized, usually using methods such as random initialization or Xavier initialization, to avoid the initial data model falling into a local optimal solution at the beginning of training, which reduces the training effect.

[0128] 3) Model training based on training samples, using the historical work orders to generate training samples to train the initial data model; during the monitoring training process, monitoring the performance, loss and training progress of the initial data model, so as to timely discover problems and improve the data structure and training method of the initial data model, based on this, through the training process of the initial data model with the training samples, the big data model is formed.

[0129] 4) Model evaluation and optimization: after the initial data model is trained using the training samples generated by the historical work orders to obtain the big data model, the performance of the trained big data model needs to be evaluated using the test data formed by the preset work orders. The performance evaluation indicators include accuracy, recall, F1 value, etc. Furthermore, according to the performance evaluation results of the big data model, the big data model is adjusted and optimized, such as adjusting hyperparameters (learning rate, batch size, etc.), adding regularization terms, etc.

[0130] 5) Model deployment and maintenance: deploy the trained big data model to the query environment. In this embodiment, cloud computing platform or container technology is used to deploy the big data model to the relevant system platform for management. After that, the performance of the big data model is regularly monitored, and logs and monitoring tools are used to detect signs of performance degradation or errors in the big data model, so as to maintain and optimize it in time.

[0131] In this embodiment, the big data model is obtained by training the initial data model. Through a reasonable training process, a high-performance big data model can be trained, thereby improving the filtering accuracy of real-time query information and accurately determining user business needs.

[0132] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the auxiliary method of the online customer service reply of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0133] The present application provides an auxiliary device for online customer service reply, and the auxiliary device for online customer service reply includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the auxiliary method for online customer service reply in the above-mentioned embodiment one.

[0134] Reference below Figure 4, which shows a schematic diagram of the structure of an auxiliary device suitable for implementing the online customer service reply of the embodiment of the present application. The auxiliary device for the online customer service reply in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 4 The auxiliary device for online customer service response shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0135] like Figure 4 As shown, the auxiliary device for online customer service reply may include a processing device 1001 (such as a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to the program stored in a read-only memory (ROM: Read Only Memory) 1002 or the program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the auxiliary device for online customer service reply are also stored. The processing device 1001, ROM1002 and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow the auxiliary device for online customer service replies to communicate wirelessly or wired with other devices to exchange data. Although the figure shows auxiliary devices for online customer service replies with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have alternatively.

[0136] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0137] The auxiliary device for online customer service replies provided by the present application adopts the auxiliary method for online customer service replies in the above embodiment, which can solve the technical problem that the quality of answers that existing online customer service can provide when receiving user inquiries is uneven, affecting the overall service experience of customers. Compared with the prior art, the beneficial effects of the auxiliary device for online customer service replies provided by the present application are the same as the beneficial effects of the auxiliary method for online customer service replies provided by the above embodiment, and the other technical features of the auxiliary device for online customer service replies are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0138] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0139] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0140] The present application provides a storage medium, which is a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, and the computer-readable program instructions are used to execute the auxiliary method for online customer service reply in the above-mentioned embodiment.

[0141] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0142] The computer-readable storage medium may be included in the auxiliary device for online customer service replies; or may exist independently without being assembled into the auxiliary device for online customer service replies.

[0143] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the auxiliary device for online customer service reply, the auxiliary device for online customer service reply implements the technical content of the auxiliary method embodiment of online customer service reply shown above.

[0144] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0145] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0146] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0147] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned auxiliary method for online customer service replies, and can solve the technical problem that the quality of answers that existing online customer service can provide when receiving user inquiries is uneven, affecting the overall service experience of customers. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the auxiliary method for online customer service replies provided by the above-mentioned embodiment, and will not be repeated here.

Claims

1. An auxiliary method for online customer service reply, characterized in that: The method includes: Get real-time inquiry information of the current work order; Filter the real-time inquiry information with a preset rule engine to obtain inquiry content corresponding to the user's business needs; The query content is input into a big data model, and knowledge points corresponding to the query content output by the big data model based on a preset knowledge base are obtained, where the knowledge points include content that solves user business needs.

2. The auxiliary method for online customer service reply as claimed in claim 1, characterized in that: After the step of obtaining the knowledge points corresponding to the query content output by the big data model based on the preset knowledge base, the method further includes: In response to the editing operation on the knowledge point, answer information corresponding to the inquiry information is determined according to the edited knowledge point and outputted.

3. The auxiliary method for online customer service reply as claimed in claim 1, characterized in that: The step of inputting the query content into a big data model and obtaining knowledge points corresponding to the query content output by the big data model based on a preset knowledge base includes: Processing the query content with the multimodal feature extraction and fusion algorithm of the big data model to obtain the user's business needs; The knowledge points are acquired from the preset knowledge base based on the user's business requirements.

4. The auxiliary method for online customer service reply as claimed in claim 3, characterized in that: The step of processing the query content with the multimodal feature extraction and fusion algorithm of the big data model to obtain the user's business needs includes: Obtaining characteristic information according to the processing result of the inquiry content; The feature information is input into the semantic network of the big data model, and the user business needs represented by the feature information are determined by the multi-level association rules of the semantic network.

5. The auxiliary method for online customer service reply as claimed in claim 1, characterized in that: The auxiliary method for online customer service reply also includes: Determine the user's business requirement at the inquiry location of the current work order; The knowledge point is marked at the inquiry position, and an association relationship between the work order and the knowledge point is created and saved.

6. The online customer service reply auxiliary method according to claim 1, characterized in that: The auxiliary method for online customer service reply also includes: When the knowledge point corresponding to the query content output by the big data model based on the preset knowledge base is not obtained, determining the user business demand of the query content; Create content that solves the user's business needs and generate new knowledge point output based on the content.

7. The online customer service reply auxiliary method according to claim 6, characterized in that: After the step of creating content that solves the user's business needs and generating knowledge point output based on the content, the process further includes: Determining whether the preset knowledge base stores a target knowledge point that is repeated with the new knowledge point; When it is determined that the target knowledge point is not stored in the preset knowledge base, storing the new knowledge point in the preset knowledge base; When it is determined that the target knowledge point is stored in the preset knowledge base, the user business demand corresponding to the new knowledge point is added to the storage content corresponding to the target knowledge point.

8. The auxiliary method for online customer service reply according to any one of claims 1 to 7, characterized in that: The auxiliary method of the online customer service reply also includes: Creating an initial data model, and training the initial data model using training samples generated by historical work orders; The large data model is obtained based on the training result of the initial data model.

9. An auxiliary device for online customer service reply, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the auxiliary method for online customer service reply as described in any one of claims 1 to 8.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the auxiliary method for online customer service reply as described in any one of claims 1 to 8 are implemented.

Images