Customer service public opinion early-warning method and apparatus, and electronic device and storage medium
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- SHANGHAI SHIZHUANG INFORMATION TECHNOLOGY CO LTD
- Filing Date
- 2025-01-06
- Publication Date
- 2026-07-09
AI Technical Summary
In the prior art, when handling consultations and complaints, the e-commerce platform customer service department has fewer considerations on public opinion prediction methods, resulting in a low prediction accuracy rate.
Work order portraits are constructed using at least one dimension of work order dimension, user dimension and order dimension, and the trained integrated learning model is used to predict the probability of public opinion occurrence, and early warning processing is carried out in combination with public opinion warning thresholds.
By constructing work order portraits in multiple dimensions, the accuracy and efficiency of public opinion warnings are improved, potential public opinion can be predicted and processed more accurately, and users' satisfaction with the platform is improved.
Smart Images
Figure CN2025070843_09072026_PF_FP_ABST
Abstract
Description
Customer service public opinion early warning method, device, electronic device and storage medium
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] This application claims priority to Chinese patent application CN202410115982.2, filed on January 26, 2024, entitled “A Customer Service Public Opinion Warning Method, Device, Electronic Device and Storage Medium,” and the entire contents of that application are incorporated herein by reference. Technical Field
[0003] The present application relates to the field of artificial intelligence technology, and more specifically, to a customer service public opinion early warning method, device, electronic device, and storage medium. Background Art
[0004] With the rapid development of the internet, the e-commerce industry has rapidly risen, and e-commerce platforms have become the main channel for consumers to shop online. However, this has also brought with it a large number of inquiries and complaints, putting enormous pressure on the customer service departments of e-commerce platforms.
[0005] However, since it is difficult for customer service staff to handle all inquiries and complaints in a timely manner, and some users may not express themselves rationally enough, which can easily lead to public opinion, related technologies have also proposed prediction methods for public opinion prediction, but they consider fewer dimensions when making public opinion predictions, resulting in low prediction accuracy. Summary of the Invention
[0006] The purpose of the embodiments of the present application is to provide a customer service public opinion warning method, device, electronic device and storage medium to improve the accuracy of customer service public opinion warning.
[0007] In a first aspect, an embodiment of the present application provides a customer service public opinion warning method, the method comprising: constructing a work order portrait based on an incoming customer service work order; wherein the work order portrait includes at least one of the public opinion occurrence probability of the work order dimension, the public opinion occurrence probability of the user dimension, and the public opinion occurrence probability of the order dimension; predicting the public opinion occurrence probability of the incoming customer service work order based on the work order portrait; and performing public opinion warning processing on the incoming customer service work order based on the public opinion occurrence probability and the public opinion warning threshold.
[0008] In the implementation process of the above solution, a work order portrait is constructed from at least one dimension among the work order dimension, user dimension and order dimension, so as to predict the probability of public opinion occurrence of the work order based on the work order portrait. This solution uses at least one dimension among the work order dimension, user dimension and order dimension to construct the work order portrait, which is conducive to improving the accuracy of public opinion warning.
[0009] In an implementation method of the first aspect, predicting the probability of public opinion occurrence of the incoming customer service work order based on the work order portrait includes: if the work order portrait contains at least two of the probability of public opinion occurrence in the work order dimension, the probability of public opinion occurrence in the user dimension, and the probability of public opinion occurrence in the order dimension, then using a trained integrated learning model to predict the probability of public opinion occurrence of the incoming customer service work order based on the work order portrait.
[0010] In the implementation process of the above solution, when the work order portrait contains the probability of public opinion occurrence in multiple dimensions, a trained integrated learning model is used to combine the probability of public opinion occurrence in multiple dimensions to jointly predict the probability of public opinion occurrence of incoming customer service work orders, which is conducive to improving the accuracy of public opinion warning of the above customer service public opinion warning method.
[0011] In an implementation method of the first aspect, constructing a work order portrait based on an incoming customer service work order includes: determining the work order subject of the incoming customer service work order; determining the work order dimension public opinion probability of the current incoming customer service work order based on the public opinion probability of the corresponding work order subject in historical work order subjects.
[0012] During the implementation of the above solution, the probability of public opinion occurring in the work order dimension of the current incoming customer service work order can be quickly determined through the work order subject of the incoming customer service work order, which is conducive to improving the public opinion warning efficiency of the above customer service public opinion warning method.
[0013] In an implementation method of the first aspect, determining the work order subject of the incoming customer service work order includes: performing word segmentation processing on the incoming customer service work order; determining the word subject of the words in the incoming customer service work order; and determining the work order subject of the incoming customer service work order based on the word subject of the words in the incoming customer service work order.
[0014] In the implementation process of the above scheme, the theme of the incoming customer service work order is determined by performing word segmentation on the incoming customer service work order and determining the word theme of the word, which is conducive to improving the accuracy of theme determination of the work order theme, and further conducive to improving the accuracy of public opinion warning of the above customer service public opinion warning method.
[0015] In an implementation method of the first aspect, determining the word topic of the words in the incoming customer service work order includes: randomly initializing the word topic of the words in the incoming customer service work order; calculating a first proportion and a second proportion for each of the words and their word topics in the incoming customer service work order; wherein the first proportion is used to indicate the proportion of words in the incoming customer service work order that are designated as the current word topic, and the second proportion is used to indicate the proportion of work order topics of customer service work orders containing the current word in historical customer service work orders that are designated as the current word topic; calculating the word topic allocation probability based on the first proportion and the second proportion; updating the word topic based on the word topic allocation probability; repeatedly iterating the above-mentioned calculation steps of the first proportion and the second proportion, the word topic allocation probability calculation step, and the word topic update step until a preset iteration stop condition is reached.
[0016] During the implementation of the above scheme, by continuously iteratively updating the word topics in the incoming customer service work orders, the work order topics of the incoming customer service work orders are determined based on the iteratively updated word topics, which is conducive to improving the accuracy of determining the topics of the work order topics, and thus is conducive to improving the accuracy of public opinion warning of the above customer service public opinion warning method.
[0017] In an implementation of the first aspect, the construction of a work order portrait based on the incoming customer service work order includes: obtaining user data associated with the incoming customer service work order; extracting user features from the user data; and using a trained user-dimensional public opinion probability prediction model to predict the user-dimensional public opinion probability based on the user features.
[0018] During the implementation of the above solution, the probability of public opinion occurrence in the user dimension is predicted by extracting user features from the user data associated with the incoming customer service work orders, thereby enriching the work order portrait, which is conducive to improving the accuracy of public opinion warning of the above customer service public opinion warning method.
[0019] In an implementation of the first aspect, constructing a work order portrait based on an incoming customer service work order includes: obtaining order data associated with the incoming customer service work order; extracting order features from the order data; and using a trained order dimension public opinion probability prediction model to predict the order dimension public opinion probability based on the order features.
[0020] During the implementation of the above solution, the probability of public opinion occurring in the order dimension is predicted by extracting order features from the order data associated with the incoming customer service work orders, thereby enriching the work order portrait, which is conducive to improving the accuracy of public opinion warning of the above customer service public opinion warning method.
[0021] In a second aspect, an embodiment of the present application provides a customer service public opinion early warning device, which includes:
[0022] A portrait construction module is used to construct a work order portrait based on the incoming customer service work order; wherein the work order portrait includes at least one of the probability of public opinion occurring in the work order dimension, the probability of public opinion occurring in the user dimension, and the probability of public opinion occurring in the order dimension;
[0023] A public opinion occurrence probability prediction module, used to predict the probability of public opinion occurrence of the incoming customer service work order based on the work order portrait;
[0024] The early warning processing module is used to perform public opinion early warning processing on the incoming customer service work order based on the probability of public opinion occurrence and the public opinion early warning threshold.
[0025] In a third aspect, an embodiment of the present application provides an electronic device comprising: a processor, a memory, and a communication bus, wherein the processor and the memory communicate with each other through the communication bus; the memory stores computer program instructions that can be executed by the processor, and when the computer program instructions are read and run by the processor, the method provided in the first aspect or any possible implementation of the first aspect is executed.
[0026] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which computer program instructions are stored. When the computer program instructions are read and run by a processor, the method provided by the first aspect or any possible implementation of the first aspect is executed.
[0027] Other features and advantages of the present application will be described in the following description and, in part, will become apparent from the description or be understood by practicing the embodiments of the present application. The objectives and other advantages of the present application can be achieved and obtained through the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.
[0029] FIG1 is a flow chart of a customer service public opinion early warning method provided in an embodiment of the present application;
[0030] FIG2 is a schematic diagram of the structure of a customer service public opinion early warning device provided in an embodiment of the present application;
[0031] FIG3 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0032] The following will describe the technical solutions in the embodiments of the present application in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present application and are therefore only examples and cannot be used to limit the scope of protection of the present application.
[0033] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.
[0034] In the description of the embodiments of this application, the technical terms "first" and "second" are used only to distinguish different objects and should not be understood to indicate or imply relative importance or implicitly specify the quantity, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, the meaning of "plurality" is more than two, unless otherwise clearly and specifically defined.
[0035] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0036] In the description of the embodiments of this application, the term "and / or" is simply a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent the following three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.
[0037] E-commerce platforms have become the primary channel for online shopping, and their customer service departments are faced with a high volume of inquiries and complaints. In this environment, user satisfaction with and evaluation of the platforms have become crucial issues. A key factor in user satisfaction with e-commerce platforms is the occurrence of public opinion. Therefore, early warnings can be used to prepare for impending public opinion, or customer service can be promptly addressed before public opinion occurs, thereby improving user satisfaction with the platform.
[0038] The public opinion prediction methods in related technologies mostly profile users and judge the probability of the user having public opinion so as to issue public opinion warnings. However, this solution has a low accuracy rate in public opinion warnings because it only considers the user dimension.
[0039] Based on this, an embodiment of the present application provides a customer service public opinion warning method. Different from the method of using user portraits to conduct public opinion warning in related technologies, the above scheme can construct a work order portrait from at least one dimension among the work order dimension, user dimension and order dimension, thereby predicting the probability of public opinion occurrence of the work order based on the work order portrait. This scheme uses at least one dimension among the work order dimension, user dimension and order dimension to construct a work order portrait, which is conducive to improving the accuracy of public opinion warning.
[0040] The following is a detailed description of the customer service public opinion warning method. Referring to FIG1 , an embodiment of the present application provides a customer service public opinion warning method, the method comprising:
[0041] Step S110: Build a work order profile based on the incoming customer service work order;
[0042] The work order profile includes at least one of the following: the probability of public opinion occurring in the work order dimension, the probability of public opinion occurring in the user dimension, and the probability of public opinion occurring in the order dimension;
[0043] Step S120: Predicting the probability of public opinion occurring for the incoming customer service ticket based on the ticket portrait;
[0044] Step S130: Perform public opinion warning processing on the incoming customer service work order based on the probability of public opinion occurrence and the public opinion warning threshold.
[0045] The work order in step S110 is a document or spreadsheet used to record and track customer issues and requests. A customer service work order is a work order created manually or automatically when a user submits an issue or request to the platform. This work order records one or more of the following: basic user information, a description of the issue or request, and the progress of the request, for use by the customer service system.
[0046] The above-mentioned customer service public opinion refers to the public opinion on the customer service department of an enterprise or organization. This public opinion usually involves customers' evaluation and opinions on the enterprise's service quality, service attitude, service efficiency, etc., and is a reflection of the customer's comprehensive attitude towards the service department.
[0047] The work order dimension public opinion occurrence probability in the above step S110 refers to the public opinion occurrence probability in a single dimension obtained from the work order dimension itself. The acquisition scheme of the work order dimension public opinion occurrence probability is as follows:
[0048] As an optional implementation method of the above-mentioned customer service public opinion warning method, the above-mentioned step S110 includes: determining the work order subject of the incoming customer service work order; determining the work order dimension public opinion probability of the current incoming customer service work order based on the public opinion occurrence probability of the corresponding work order subject in the historical work order subject.
[0049] It can be understood that the probability of public opinion occurring in the work order dimension of the above-mentioned current incoming customer service work order can be determined by the proportion of work orders with public opinion occurring under the corresponding work order topic to the total number of historical work orders under the corresponding topic to determine the probability of public opinion occurring in the corresponding work order topic in the historical work order topic.
[0050] It is understandable that the above scheme can use a rule-based topic extraction method to extract the ticket topics of incoming customer service tickets. The rule-based topic extraction method is such as keyword matching, regular expressions, etc. The rules can be manually set extraction rules.
[0051] The above solution can quickly determine the probability of public opinion occurring in the work order dimension of the current incoming customer service work order through the work order subject of the incoming customer service work order, which is conducive to improving the public opinion warning efficiency of the above customer service public opinion warning method.
[0052] In addition, it is understandable that the above-mentioned work order topic can also be determined using a reinforcement learning-based method. An optional implementation of the reinforcement learning-based topic extraction method is introduced below:
[0053] As an optional implementation method of the above-mentioned customer service public opinion warning method, the above-mentioned determination of the work order theme of the incoming customer service work order includes: performing word segmentation processing on the incoming customer service work order; determining the word theme of the words in the incoming customer service work order; and determining the work order theme of the incoming customer service work order based on the word theme of the words in the incoming customer service work order.
[0054] It is understandable that a mature word segmentation model can be used to perform word segmentation processing on incoming customer service tickets.
[0055] The word topics of the words in the above-mentioned incoming customer service work orders and the work order topics of the incoming customer service work orders can be obtained through word segmentation algorithms such as the NMF (Non-negative Matrix Factorization) algorithm, the LSA (Latent Semantic Analysis, Singular Value Decomposition) algorithm, and the HDP (Hierarchical Dirichlet Process) algorithm. The above-mentioned NMF algorithm, LSA algorithm, and HDP algorithm are all relatively mature algorithms in the relevant technology and will not be repeated in the implementation of this application.
[0056] The above scheme determines the subject of the incoming customer service work order by performing word segmentation on the incoming customer service work order and determining the word subject of the word, which is conducive to improving the accuracy of the subject determination of the work order subject, and further helps to improve the accuracy of the public opinion warning of the above customer service public opinion warning method.
[0057] As an optional implementation method of the above-mentioned customer service public opinion warning method, the above-mentioned determination of the word theme of the words in the incoming customer service work order includes: randomly initializing the word theme of the words in the incoming customer service work order; calculating the first proportion and the second proportion for each word and its word theme in the incoming customer service work order; wherein the first proportion is used to indicate the proportion of words in the incoming customer service work order that are designated as the current word theme, and the second proportion is used to indicate the proportion of work order themes of customer service work orders containing the current word in historical customer service work orders that are designated as the current word theme; calculating the word theme allocation probability based on the first proportion and the second proportion; updating the word theme based on the word theme allocation probability; repeating the above-mentioned calculation steps of the first proportion and the second proportion, the word theme allocation probability calculation step, and the word theme update step until the preset iteration stop condition is reached. This implementation method is, for example:
[0058] For incoming customer service tickets after word segmentation, the word topic of each word is randomly initialized to one of K topics (the K topics can be all optional ticket topics selected in advance);
[0059] For an incoming customer service ticket D and a word W in the incoming customer service ticket, calculate a first proportion and a second proportion. The first proportion indicates the proportion of words in the incoming customer service ticket D that are assigned to the word topic of the current word W, and the second proportion indicates the probability of all historical customer service tickets containing the word W being assigned to the ticket topic T.
[0060] The calculation method of the first proportion is: P1(T|D)
[0061] The calculation method of the second proportion is: P2(W|T)
[0062] The method for calculating the word topic assignment probability P3 based on the first proportion and the second proportion is: P3 = P1(T|D)*P2(W|T)
[0063] Update the word topic according to the above word topic assignment probability;
[0064] The first proportion calculation step, the second proportion calculation step, the word topic allocation probability calculation step, and the word topic update step are re-iterated until the word topic no longer changes, or until a preset number of re-iterations are performed.
[0065] It is understandable that the work order topic calculated by the above solution may be an LDA (Latent Dirichlet Allocations) topic.
[0066] The above scheme continuously iterates and updates the word topics in the incoming customer service work orders, thereby determining the work order topics of the incoming customer service work orders based on the iteratively updated word topics, which is conducive to improving the accuracy of determining the topics of the work order topics, and further helps to improve the accuracy of public opinion warning of the above customer service public opinion warning method.
[0067] It is understandable that before performing word segmentation on the incoming customer service work order, the incoming customer service work order may also be pre-processed by performing content cleaning, removing special characters and punctuation marks, and so on.
[0068] The following is a detailed introduction to the method for obtaining the probability of public opinion occurrence in the user dimension:
[0069] As an optional implementation method of the above-mentioned customer service public opinion warning method, the above-mentioned step S110 includes: obtaining user data associated with the incoming customer service work order; extracting user features from the user data; using a trained user-dimensional public opinion probability prediction model to predict the user-dimensional public opinion probability based on user features.
[0070] It is understandable that incoming customer service tickets will be associated with users, so user data can be obtained based on the users associated with the incoming customer service tickets. User data includes age, region, and historical transaction behavior, where historical transaction behavior includes but is not limited to placing orders, after-sales service, incoming calls, complaints, etc.
[0071] In addition, it is understandable that after obtaining the user data associated with the incoming customer service work order, the user data may be pre-processed such as missing value processing and feature binning processing.
[0072] The above-mentioned user-dimensional public opinion probability prediction model can adopt the LightGBM model. The LightGBM model is a gradient boosting framework model based on the decision tree algorithm. The LightGBM model can be trained based on the existing user-level public opinion black samples, so that the LightGBM model can predict the probability of user-dimensional public opinion according to user characteristics.
[0073] The above scheme predicts the probability of user-dimensional public opinion by extracting user features from user data associated with incoming customer service work orders, thereby enriching the work order portrait, which is conducive to improving the accuracy of public opinion warning of the above customer service public opinion warning method.
[0074] The following describes the method for obtaining the public opinion probability of the order dimension:
[0075] As an optional implementation method of the above-mentioned customer service public opinion warning method, the above-mentioned step S110 includes: obtaining order data associated with the incoming customer service work order; extracting order features from the order data; using a trained order dimension public opinion probability prediction model to predict the order dimension public opinion probability based on the order features.
[0076] It is understandable that incoming customer service work orders will also be associated with orders, so order data can be obtained based on the orders associated with the incoming customer service work orders. Order data includes but is not limited to current transaction orders, historical transaction orders, product types, product prices, logistics time, etc.
[0077] In addition, it is understandable that after obtaining the order data associated with the incoming customer service work order, the order data can also be pre-processed by filling missing values and performing feature binning.
[0078] The above-mentioned order-dimensional public opinion probability prediction model can also adopt the LightGBM model. The LightGBM model can be trained based on the existing order-level public opinion black samples, so that the LightGBM model can predict the probability of user-dimensional public opinion based on order characteristics.
[0079] The above solution predicts the probability of public opinion occurring in the order dimension by extracting order features from the order data associated with the incoming customer service work orders, thereby enriching the work order portrait, which is conducive to improving the accuracy of public opinion warning of the above customer service public opinion warning method.
[0080] The following is a detailed description of step S120:
[0081] Since a work order profile can include at least one of the following: the probability of public opinion occurring in the work order dimension, the probability of public opinion occurring in the user dimension, and the probability of public opinion occurring in the order dimension, the following describes in detail two implementation methods: one in which the work order profile only includes one public opinion probability, and the other in which it includes multiple public opinion probability.
[0082] In the first implementation, the work order profile only includes one item, the probability of public opinion occurrence. In this case, step S120 includes:
[0083] The probability of public opinion occurrence included in the above work order portrait is determined as the probability of public opinion occurrence of the incoming customer service work order.
[0084] In a second embodiment, the work order profile includes multiple public opinion occurrence probabilities. In this case, step S120 includes:
[0085] Using the trained ensemble learning model, the probability of public opinion occurring in incoming customer service tickets is predicted based on the ticket portrait.
[0086] Ensemble learning refers to the idea of combining individual machine learning algorithms for modeling. Ensemble learning includes three methods: bagging (bootstrap aggregation), boosting, and stacking. The core idea of bagging is collective voting, where each base model has only one vote, and the result with the most votes is the final result. The core idea of boosting is to select elite models, test and screen the base models, give elite models more votes, give other models fewer votes, and then combine the votes of all models to obtain the final result. Stacking consists of two parts: a meta-estimator and a base estimator. The meta-estimator uses the predictions of the base estimators as input variables and ultimately makes a prediction for the explained variable. The base estimator can be composed of different algorithmic models (such as logistic regression and k-NN), and the meta-estimator can also choose a completely different algorithmic model (such as a decision tree) to achieve better results in reducing error.
[0087] The integrated learning model in the embodiment of the present application can be learned in a Boosting manner. For example, by combining historical public opinion samples, the influence relationship between the probability of public opinion occurrence in the work order dimension, the probability of public opinion occurrence in the user dimension, and the probability of public opinion occurrence in the order dimension on the probability of public opinion occurrence of the incoming customer service work order is determined respectively. For example, it is determined that the probability of public opinion occurrence in the work order dimension has a greater influence on the probability of public opinion occurrence in the incoming customer service work order, while the probability of public opinion occurrence in the user dimension and the probability of public opinion occurrence in the order dimension have a smaller influence on the probability of public opinion occurrence in the incoming customer service work order. Therefore, the voting right with a greater probability of public opinion occurrence in the work order dimension can be given, while the voting right with a smaller probability of public opinion occurrence in the user dimension and the probability of public opinion occurrence in the order dimension can be given. Based on this, the integrated learning method shown in the following content can be adopted:
[0088] (1) When the probability of public opinion occurring in the work order dimension is high, if the probability of public opinion occurring in the user dimension is high, or the probability of public opinion occurring in the order dimension is high, or the probability of public opinion occurring in the user dimension is medium and the probability of public opinion occurring in the order dimension is medium, the ensemble learning model assigns a higher value to the probability of public opinion occurring;
[0089] (2) When the probability of public opinion occurring in the work order dimension is medium, if the probability of public opinion occurring in the user dimension is high and the probability of public opinion occurring in the order dimension is high, the ensemble learning model assigns a higher value to the probability of public opinion occurring;
[0090] (3) When the probability of public opinion occurring in the work order dimension is low, the integrated learning model assigns a lower value to the probability of public opinion occurring.
[0091] When the work order portrait contains the probability of public opinion occurrence in multiple dimensions, the above scheme uses a trained integrated learning model to combine the probability of public opinion occurrence in multiple dimensions to jointly predict the probability of public opinion occurrence of incoming customer service work orders, which is conducive to improving the accuracy of public opinion warning of the above customer service public opinion warning method.
[0092] The following is a detailed description of step S130:
[0093] It is understandable that the public opinion warning threshold in step S130 can be a preset public opinion warning threshold, or it can be an optimal public opinion warning threshold selected regularly or in real time based on the risk work order recall rate and accuracy rate as evaluation indicators.
[0094] The embodiment of the present application also provides a specific implementation of the above-mentioned customer service public opinion warning method in a certain scenario. In this scenario, the work order profile includes the public opinion occurrence probability of three dimensions: the work order dimension public opinion occurrence probability, the user dimension public opinion occurrence probability, and the order dimension public opinion occurrence probability. In this case, the above-mentioned customer service public opinion warning method may include:
[0095] Step 1: Data collection;
[0096] Obtain incoming customer service tickets and the order data and user data associated with them;
[0097] Step 2: Sort out work order dimension data;
[0098] Clean the content of incoming customer service tickets, remove special characters and punctuation marks, and perform word segmentation on them.
[0099] Step 3: Random initialization of word topics;
[0100] For incoming customer service tickets, the word topic of each word is randomly initialized to one of K topics (K topics are pre-selected optional topics);
[0101] Step 4: Word theme iteration;
[0102] For an incoming customer service ticket D, browse word W and calculate P(T|D) and P(W|T). P(T|D) represents the proportion of words assigned to topic T in the incoming customer service ticket D; P(W|T) represents the proportion of all historical customer service tickets containing word W that are assigned to topic T.
[0103] Consider all other words and their topic assignments, and reassign word W to topic T with probability P(T|D)*P(W|T);
[0104] Step 5: LDA topic model establishment;
[0105] Repeat step 4 above until the word theme stabilizes, and determine the ticket theme of the incoming customer service ticket based on the word theme of the words in the incoming customer service ticket;
[0106] Step 6: Calculate the probability of public opinion occurring in the work order dimension;
[0107] The probability of public opinion occurring in the ticket dimension is determined based on the ticket subject of the incoming customer service ticket and the probability of public opinion occurring in the corresponding ticket subject in historical customer service tickets.
[0108] Step 7: Sorting out user dimension data;
[0109] User data includes but is not limited to: age, region, and historical transaction behavior;
[0110] Among them, historical transaction behaviors include but are not limited to: order placement, after-sales service, incoming calls, complaints, etc.;
[0111] It is understandable that after combing the data, the data can also be preprocessed by filling missing values and feature binning;
[0112] Step 8: Calculate the probability of public opinion occurring in the user dimension;
[0113] Based on the LightGBM supervised model trained with the existing user-dimensional public opinion black samples, the probability of user-dimensional public opinion occurrence is predicted;
[0114] Step 9: Sort out order dimension data;
[0115] Order data includes but is not limited to: current transaction order information, historical transaction order information, product type, product price, logistics time, etc.;
[0116] It is understandable that after combing the data, the data can also be preprocessed by filling missing values and feature binning;
[0117] Step 10: Calculate the probability of public opinion occurring in the order dimension;
[0118] The LightGBM supervised model trained based on the existing order-dimensional public opinion black samples predicts the probability of order-dimensional public opinion occurrence;
[0119] Step 11: Real-time integration of customer service ticket features;
[0120] Based on the unique ID of the real-time customer service ticket, the ticket profile of the incoming customer service ticket is associated and the entire profile is converted into an integrated learning feature;
[0121] Step 12: Real-time model training;
[0122] Based on existing sample data, through ensemble learning, the recall rate and precision rate of risk tickets are used as evaluation indicators to determine the optimal public opinion warning threshold;
[0123] Step 13: Based on the optimal public opinion warning threshold, issue public opinion warnings for incoming customer service tickets in real time, and take timely action on customer service tickets that have received warnings.
[0124] It should be pointed out that the user-related data mentioned in the above content is obtained with the user's consent.
[0125] Referring to FIG. 2 , based on the same inventive concept, an embodiment of the present application further provides a customer service public opinion early warning device 200 , which includes:
[0126] A portrait construction module 210 is configured to construct a work order portrait based on an incoming customer service work order; wherein the work order portrait includes at least one of the probability of public opinion occurring in the work order dimension, the probability of public opinion occurring in the user dimension, and the probability of public opinion occurring in the order dimension;
[0127] The public opinion occurrence probability prediction module 220 is used to predict the public opinion occurrence probability of the incoming customer service work order based on the work order portrait;
[0128] The warning processing module 230 is used to perform public opinion warning processing on the incoming customer service work order based on the probability of public opinion occurrence and the public opinion warning threshold.
[0129] As an optional implementation of the above-mentioned customer service public opinion early warning device, the above-mentioned public opinion occurrence probability prediction module 220 is specifically used for: when the work order portrait contains at least two of the work order dimension public opinion occurrence probability, the user dimension public opinion occurrence probability and the order dimension public opinion occurrence probability, using the trained integrated learning model, according to the work order portrait, predicting the public opinion occurrence probability of the incoming customer service work order.
[0130] As an optional implementation of the customer service public opinion early warning device, the portrait building module 210 includes:
[0131] The work order subject determination unit is used to determine the work order subject of the incoming customer service work order;
[0132] The work order dimension public opinion occurrence probability determination unit determines the work order dimension public opinion occurrence probability of the current incoming customer service work order based on the public opinion occurrence probability of the corresponding work order topic in the historical work order topics.
[0133] As an optional implementation method of the above-mentioned customer service public opinion warning device, the above-mentioned work order theme determination unit is specifically used to: perform word segmentation processing on the incoming customer service work order; determine the word theme of the words in the incoming customer service work order; and determine the work order theme of the incoming customer service work order based on the word theme of the words in the incoming customer service work order.
[0134] As an optional implementation manner of the above-mentioned customer service public opinion early warning device, the above-mentioned determination of the word theme of the words in the incoming customer service work order includes: randomly initializing the word theme of the words in the incoming customer service work order; calculating a first proportion and a second proportion for each of the words and the word theme in the incoming customer service work order; wherein, the first proportion is used to indicate the proportion of words in the incoming customer service work order designated as the current word theme, and the second proportion is used to indicate the proportion of work order themes of customer service work orders containing the current word in historical customer service work orders that are designated as the current word theme; calculating the word theme allocation probability based on the first proportion and the second proportion; updating the word theme based on the word theme allocation probability; repeatedly iterating the above-mentioned calculation steps of the first proportion and the second proportion, the word theme allocation probability calculation step, and the word theme update step until the preset iteration stop condition is reached.
[0135] As an optional implementation of the customer service public opinion early warning device, the portrait building module 210 includes:
[0136] A user data acquisition unit, configured to acquire user data associated with the incoming customer service work order;
[0137] A user feature extraction unit, configured to extract user features from the user data;
[0138] The user-dimensional public opinion occurrence probability prediction unit is used to use the trained user-dimensional public opinion occurrence probability prediction model to predict the user-dimensional public opinion occurrence probability according to the user characteristics.
[0139] As an optional implementation of the customer service public opinion early warning device, the portrait building module 210 includes:
[0140] An order data acquisition unit, configured to acquire order data associated with the incoming customer service work order;
[0141] An order feature extraction unit, configured to extract order features from the order data;
[0142] The order dimension public opinion occurrence probability prediction unit is used to use the trained order dimension public opinion occurrence probability prediction model to predict the order dimension public opinion occurrence probability according to the order characteristics.
[0143] FIG3 is a schematic diagram of an electronic device provided in an embodiment of the present application. Referring to FIG3 , the electronic device 300 includes: a processor 310, a memory 320, and a communication interface 330. These components are interconnected and communicate with each other via a communication bus 340 and / or other forms of connection mechanisms (not shown).
[0144] The memory 320 includes one or more (only one is shown in the figure), which may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc. The processor 310 and other possible components can access the memory 320 and read and / or write data therein.
[0145] The processor 310 includes one or more (only one is shown in the figure), which can be an integrated circuit chip with signal processing capabilities. The above-mentioned processor 310 can be a general-purpose processor, including a central processing unit (CPU), a microcontroller unit (MCU), a network processor (NP), or other conventional processors; it can also be a special-purpose processor, including a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.
[0146] Communication interface 330 includes one or more (only one is shown in the figure) interfaces that can be used to communicate directly or indirectly with other devices to exchange data. For example, communication interface 330 can be an Ethernet interface; a mobile communication network interface, such as a 3G, 4G, or 5G network interface; or other types of interfaces that have data transmission and reception capabilities.
[0147] One or more computer program instructions can be stored in the memory 320, and the processor 310 can read and run these computer program instructions to implement the customer service public opinion warning method and other desired functions provided in the embodiment of the present application.
[0148] It will be understood that the structure shown in FIG3 is for illustration only, and the electronic device 300 may also include more or fewer components than shown in FIG3 , or have a configuration different from that shown in FIG3 . The components shown in FIG3 may be implemented using hardware, software, or a combination thereof. For example, the electronic device 300 may be a single server (or other device with computing processing capabilities), a combination of multiple servers, a cluster of a large number of servers, etc., and may be either a physical device or a virtual device.
[0149] The present application also provides a computer-readable storage medium having computer program instructions stored thereon. When the computer program instructions are read and executed by a computer processor, the computer program instructions execute the customer service public opinion early warning method provided in the present application. For example, the computer-readable storage medium can be implemented as memory 320 in electronic device 300 in FIG. 3 .
[0150] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interface, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0151] In addition, the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0152] Furthermore, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0153] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.
Claims
1. A customer service public opinion early warning method, characterized in that: The method comprises: Build a work order profile based on incoming customer service work orders; wherein the work order profile includes at least one of the probability of public opinion occurring in the work order dimension, the probability of public opinion occurring in the user dimension, and the probability of public opinion occurring in the order dimension; Predicting the probability of public opinion occurring for the incoming customer service ticket based on the ticket portrait; According to the probability of occurrence of public opinion and the public opinion warning threshold, public opinion warning processing is performed on the incoming customer service work order.
2. The customer service public opinion early warning method according to claim 1 is characterized in that: The predicting, based on the work order portrait, the probability of public opinion occurring for the incoming customer service work order includes: If the work order portrait includes at least two of the probability of public opinion occurring in the work order dimension, the probability of public opinion occurring in the user dimension, and the probability of public opinion occurring in the order dimension, the trained integrated learning model is used to predict the probability of public opinion occurring in the incoming customer service work order based on the work order portrait.
3. The customer service public opinion early warning method according to claim 1, characterized in that: The process of constructing a work order profile based on incoming customer service work orders includes: Determine the ticket subject of the incoming customer service ticket; According to the probability of public opinion occurrence of the corresponding work order topic in the historical work order topics, the probability of public opinion occurrence of the work order dimension of the current incoming customer service work order is determined.
4. The customer service public opinion early warning method according to claim 3 is characterized in that: Determining the subject of the incoming customer service ticket includes: Perform word segmentation on the incoming customer service work order; Determining a word theme of a word in the incoming customer service ticket; Determine a work order subject of the incoming customer service work order based on the word subject of the words in the incoming customer service work order.
5. The customer service public opinion early warning method according to claim 4 is characterized in that: Determining the word theme of the words in the incoming customer service work order includes: Randomly initializing the word topics of the words in the incoming customer service work order; For each of the words and the word topics in the incoming customer service work order, a first proportion and a second proportion are calculated; wherein the first proportion is used to indicate the proportion of words in the incoming customer service work order that are assigned to the current word topic, and the second proportion is used to indicate the proportion of historical customer service work orders that contain the current word and whose work order topics are assigned to the current word topic; Calculating a word topic assignment probability based on the first proportion and the second proportion; Update the word topic according to the word topic assignment probability; Repeat the above-mentioned steps of calculating the first proportion and the second proportion, calculating the word topic assignment probability, and updating the word topic until a preset iteration stopping condition is reached.
6. The customer service public opinion early warning method according to any one of claims 1 to 5, characterized in that: The process of constructing a work order profile based on incoming customer service work orders includes: Obtaining user data associated with the incoming customer service ticket; extracting user features from the user data; The trained user-dimensional public opinion occurrence probability prediction model is used to predict the user-dimensional public opinion occurrence probability based on the user characteristics.
7. The customer service public opinion early warning method according to any one of claims 1 to 5, characterized in that: The process of constructing a work order profile based on incoming customer service work orders includes: Obtaining order data associated with the incoming customer service work order; extracting order features from the order data; The trained order-dimension public opinion occurrence probability prediction model is used to predict the order-dimension public opinion occurrence probability based on the order characteristics.
8. A customer service public opinion early warning device, characterized in that: The device comprises: A portrait construction module is used to construct a work order portrait based on the incoming customer service work order; wherein the work order portrait includes at least one of the probability of public opinion occurring in the work order dimension, the probability of public opinion occurring in the user dimension, and the probability of public opinion occurring in the order dimension; A public opinion occurrence probability prediction module, used to predict the probability of public opinion occurrence of the incoming customer service work order based on the work order portrait; The early warning processing module is used to perform public opinion early warning processing on the incoming customer service work order based on the probability of public opinion occurrence and the public opinion early warning threshold.
9. An electronic device, characterized in that: include: A processor, a memory and a communication bus, wherein the processor and the memory communicate with each other via the communication bus; The memory stores program instructions that can be executed by the processor, and the processor can execute the method according to any one of claims 1 to 7 by calling the program instructions.
10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, which, when executed by a computer, enable the computer to perform the method according to any one of claims 1 to 7.