User demand processing method and device

Abstract

The invention discloses a user appeal processing method and device. The method comprises the steps that in response to user appeal information of a target user, multiple information keywords are extracted through analysis, and the appeal level of the target user is evaluated and determined; selecting matched appeal features and intentions from a preset appeal information feature library and a preset appeal intention library based on the appeal levels; extracting a plurality of target user information features by using the target user portrait; candidate features matched with the information keywords are screened from an appeal information feature library, and target appeal information features matched with the user information features are further recognized; and positioning a corresponding appeal intention from the appeal intention library according to the target appeal information features, generating personalized appeal feedback information, and directly pushing the personalized appeal feedback information to the target user. The technical problems that traditional manual customer service is low in user demand processing efficiency, poor in accuracy and difficult to meet user requirements are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of customer service management, in particular to a user appeal processing method and device. BACKGROUND

[0002] Under the background of the digital era, the interaction between enterprises and consumers is increasingly frequent, generating a large amount of data, especially user feedback data, including but not limited to complaints, consultations, suggestions, etc. These feedback information is a key resource for enterprises to improve products and services and enhance user experience.

[0003] However, the traditional feedback processing mechanism, especially in the field of customer service, mainly relies on manual processing. With the increase of user base and the surge of feedback quantity, the manual processing method gradually reveals its inherent limitations. On the one hand, facing a large number of user work orders, manually reading, understanding and replying to each work order not only consumes time and effort, but also may lead to low processing efficiency and response delay, affecting user experience. On the other hand, due to the differences in personal experience and technical level of customer service personnel, the same type of user feedback may be interpreted and processed differently, which reduces the consistency and satisfaction of the processing results. Especially in the processing of complex user feedback, the lack of effective use of user's personalized characteristics often makes the reply information too generalized, making it difficult to accurately solve the user's specific problems.

[0004] In recent years, the development of artificial intelligence technology has provided new solutions for enterprises. Through natural language processing, machine learning and other technologies, enterprises have begun to try to automatically process user work orders to reduce the work burden of customer service personnel and improve processing efficiency. Although it can achieve automatic processing to a certain extent, this method is limited to the processing of new work orders, and the understanding of user demand depends on simple emotion analysis, which may cause deviation in understanding user demand in complex scenarios, affecting the subsequent processing effect.

[0005] In view of the above problems, no effective solution has been proposed so far. SUMMARY

[0006] The embodiments of the present application provide a user appeal processing method and device to at least solve the technical problems of low efficiency, poor accuracy and difficulty in meeting user demand of traditional manual customer service in processing user appeal.

[0007] According to an aspect of the embodiments of the present application, a user appeal processing method is provided, comprising: in response to user appeal information of a target user, extracting a plurality of information keywords in the user appeal information, and determining a target appeal level of the target user according to the plurality of information keywords; determining a target appeal information feature library and a target appeal intent library corresponding to the target appeal level, wherein the target appeal information feature library stores a plurality of appeal information features, and the target appeal intent library stores a plurality of appeal intents corresponding to different appeal information features; obtaining a target user portrait of the target user, and extracting a plurality of target user information features from the target user portrait; determining a plurality of candidate appeal information features matched with the plurality of information keywords from the target appeal information feature library, and determining a target appeal information feature matched with the plurality of target user information features from the plurality of candidate appeal information features; determining a target appeal intent corresponding to the target appeal information feature from the target appeal intent library, and generating appeal feedback information according to the target appeal intent, and feeding back the appeal feedback information to the target user.

[0008] Optionally, the target appeal level of the target user is determined according to the plurality of information keywords, comprising: performing part-of-speech tagging on the plurality of information keywords respectively to obtain part-of-speech tagging results; performing sentiment analysis on the plurality of information keywords respectively to obtain sentiment analysis results; constructing a keyword feature vector corresponding to the user appeal information according to the part-of-speech tagging results and the sentiment analysis results; calculating a first similarity between the keyword feature vector and a preset feature vector; determining the target appeal level corresponding to the first similarity from a preset appeal level mapping table, wherein the appeal level mapping table stores a mapping relationship between different similarities and different appeal levels.

[0009] Optionally, the target appeal information feature library and the target appeal intent library corresponding to the target appeal level are determined, comprising: determining the target appeal information feature library corresponding to the target appeal level from a preset appeal information feature library mapping table, wherein the appeal information feature library mapping table stores a mapping relationship between different appeal levels and different appeal information feature libraries; determining the target appeal intent library corresponding to the target appeal level from a preset appeal intent library mapping table, wherein the appeal intent library mapping table stores a mapping relationship between different appeal levels and different appeal intent libraries.

[0010] Optionally, the target user portrait of the target user is acquired, including: acquiring user data of the target user, and pre-processing the user data, wherein the pre-processing includes at least one of the following: data cleaning, standardization processing; analyzing the pre-processed user data by using a pre-trained user classification model, extracting target user features corresponding to the target user, and determining a target user type of the target user; determining a target demand scene according to user demand information; determining a target standard user portrait corresponding to the target user type and the target demand scene from a preset user portrait library, wherein the user portrait library stores a mapping relationship between different user types, different demand scenes and different standard user portraits, and each standard user portrait includes standard user features and corresponding standard portrait parameters of a type of user in a demand scene; adjusting the target standard user portrait according to the target user features to obtain the target user portrait of the target user.

[0011] Optionally, the target user portrait of the target user is acquired, including: acquiring user data of the target user, and pre-processing the user data, wherein the pre-processing includes at least one of the following: data cleaning, standardization processing; analyzing the pre-processed user data by using a pre-trained user classification model, extracting target user features corresponding to the target user, and determining a target user type of the target user; determining a target demand scene according to user demand information; determining a target standard user portrait corresponding to the target user type and the target demand scene from a preset user portrait library, wherein the user portrait library stores a mapping relationship between different user types, different demand scenes and different standard user portraits, and each standard user portrait includes standard user features and corresponding standard portrait parameters of a type of user in a demand scene; adjusting the target standard user portrait according to the target user features to obtain the target user portrait of the target user.

[0012] Optionally, the target user portrait of the target user is acquired, including: acquiring user data of the target user, and pre-processing the user data, wherein the pre-processing includes at least one of the following: data cleaning, standardization processing; analyzing the pre-processed user data by using a pre-trained user classification model, extracting target user features corresponding to the target user, and determining a target user type of the target user; determining a target demand scene according to user demand information; determining a target standard user portrait corresponding to the target user type and the target demand scene from a preset user portrait library, wherein the user portrait library stores a mapping relationship between different user types, different demand scenes and different standard user portraits, and each standard user portrait includes standard user features and corresponding standard portrait parameters of a type of user in a demand scene; adjusting the target standard user portrait according to the target user features to obtain the target user portrait of the target user.

[0013] Optionally, the target user portrait of the target user is acquired, including: acquiring user data of the target user, and pre-processing the user data, wherein the pre-processing includes at least one of the following: data cleaning, standardization processing; analyzing the pre-processed user data by using a pre-trained user classification model, extracting target user features corresponding to the target user, and determining a target user type of the target user; determining a target demand scene according to user demand information; determining a target standard user portrait corresponding to the target user type and the target demand scene from a preset user portrait library, wherein the user portrait library stores a mapping relationship between different user types, different demand scenes and different standard user portraits, and each standard user portrait includes standard user features and corresponding standard portrait parameters of a type of user in a demand scene; adjusting the target standard user portrait according to the target user features to obtain the target user portrait of the target user.

[0014] Optionally, the determining, from the plurality of candidate appeal information features, a target appeal information feature matching the plurality of target user information features comprises: for each candidate appeal information feature, calculating a correlation degree between the candidate appeal information feature and each target user information feature respectively, and taking an average of the plurality of correlation degrees as a comprehensive correlation degree between the candidate appeal information feature and the target user, wherein the correlation degree comprises one of a cosine similarity and a Pearson correlation coefficient; and determining the candidate appeal information feature with the maximum comprehensive correlation degree with the target user as the target appeal information feature.

[0015] Optionally, the generating, according to the target appeal intention, the appeal feedback information comprises: determining, from a preset output template library, a target output template corresponding to the target appeal intention, wherein the output template library stores a mapping relationship between different appeal intentions and different output templates; determining, from a preset output corpus library, a target output corpus corresponding to the target output template and the target user portrait, wherein the output corpus library stores a mapping relationship between different output templates, different user portraits and different output corpora; and filling the target output corpus into the target output template to obtain the appeal feedback information.

[0016] Optionally, after the appeal feedback information is fed back to the target user, the method further comprises: determining a target user group corresponding to the target user type; and pushing the appeal feedback information to all users in the target user group.

[0017] According to another aspect of the embodiments of the present application, a user appeal processing apparatus is further provided, comprising: a level determination module configured to extract a plurality of information keywords from user appeal information of a target user, and determine a target appeal level of the target user according to the plurality of information keywords; a library selection module configured to determine a target appeal information feature library and a target appeal intention library corresponding to the target appeal level, wherein the target appeal information feature library stores a plurality of appeal information features, and the target appeal intention library stores a plurality of appeal intentions corresponding to different appeal information features; a user feature determination module configured to obtain a target user portrait of the target user, and extract a plurality of target user information features from the target user portrait; an appeal feature determination module configured to determine a plurality of candidate appeal information features matching the plurality of information keywords from the target appeal information feature library, and determine a target appeal information feature matching the plurality of target user information features from the plurality of candidate appeal information features; and an appeal feedback module configured to determine a target appeal intention corresponding to the target appeal information feature from the target appeal intention library, generate appeal feedback information according to the target appeal intention, and feed back the appeal feedback information to the target user.

[0018] According to another aspect of the embodiments of the present application, a computer program product is also provided, which comprises a computer program, wherein the computer program, when executed by a processor, implements the user appeal processing method described above.

[0019] According to another aspect of the embodiments of the present application, an electronic device is also provided, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the user appeal processing method described above through the computer program.

[0020] In the embodiments of the present application, by extracting information keywords from user appeal information, the core of user appeal is quickly identified, avoiding the possible interpretation bias of artificial customer service, and improving the speed and accuracy of problem positioning; by establishing a user appeal information feature library and a user appeal intent library associated with the appeal level, the system can call appropriate information features and intents according to the specific content of the user appeal, and form a structured processing flow. This not only reduces the task burden of artificial customer service, but also makes the system response more standardized and accurate, avoiding the inconsistency of service caused by individual differences of customer service. Combined with user portrait information, personalized response is made to meet user needs; through the matching of keywords and information feature library, the system can accurately identify the essence of user appeal, and further refine it combined with user information features, improving the problem processing pertinence; after determining the target appeal intent, the system can automatically generate feedback information according to the preset template or algorithm, not only improving the speed of information transmission, but also ensuring the professionalism and consistency of feedback information, thereby solving the technical problems of low efficiency, poor accuracy and difficulty in meeting user needs in traditional artificial customer service processing of user appeal. BRIEF DESCRIPTION OF DRAWINGS

[0021] The accompanying drawings described herein are used to provide further understanding of the present application, and form a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application, and do not constitute improper limitations on the present application. In the drawings:

[0022] Figure 1 is a flowchart of an optional user appeal processing method according to an embodiment of the present application;

[0023] Figure 2 is a flowchart of an optional target appeal level determination method according to an embodiment of the present application;

[0024] Figure 3 is a flowchart of an optional target user portrait acquisition method according to an embodiment of the present application;

[0025] Figure 4 is a structural diagram of an optional user appeal processing device according to an embodiment of the present application;

[0026] Figure 5 is a structural schematic diagram of an optional electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to enable persons skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by persons skilled in the art without creative work should fall within the protection scope of the present application.

[0028] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0029] In order to better understand the embodiments of the present application, first, the following translation explanation is given for some nouns or terms appearing in the description of the embodiments of the present application:

[0030] Cosine similarity: a measure of the angle between two non-zero vectors, used to calculate the degree of similarity between them. In a multi-dimensional space, the cosine value of two vectors is equal to their dot product divided by the product of their module lengths. The value range of cosine similarity is between -1 and 1, where 1 indicates that the two vectors are exactly the same (angle is 0 degrees), -1 indicates that the two vectors are completely opposite (angle is 180 degrees), and 0 indicates that the two vectors are orthogonal (angle is 90 degrees, i.e. no similarity). Cosine similarity is widely used in information retrieval, natural language processing, recommendation systems, etc. to compare the similarity of texts or data.

[0031] Silhouette Coefficient: A quality metric used to evaluate the results of clustering, which combines the measures of cohesion (how tightly the samples within a cluster are) and separation (how far the samples of different clusters are) of the clusters. The Silhouette Coefficient has a value range between -1 and 1, with values closer to 1 indicating that the samples are closer to their own cluster and further away from other clusters, suggesting good clustering results; values closer to -1 indicate the opposite, with samples closer to other clusters; values close to 0 indicate that the samples are near the boundary of their cluster. The Silhouette Coefficient can be used to select the optimal number of clusters or evaluate the performance of different clustering algorithms.

[0032] Elbow Rule: A method used to determine the optimal number of clusters in cluster analysis. This method is based on K-means clustering analysis, by calculating the Within-Cluster Sum of Squares (WCSS) under different cluster numbers to determine the most suitable cluster number. Generally, as the number of clusters increases, the WCSS gradually decreases, but when reaching a certain cluster number, the reduction of WCSS will significantly decrease, forming an "elbow" shape, and this point is called the "elbow point", which is usually considered as the optimal number of clusters, because increasing more clusters cannot significantly improve the clustering quality, but may introduce overfitting problems.

[0033] Pearson Correlation Coefficient: A statistical measure of the degree of linear correlation between two variables, with a value range between -1 and 1. 1 indicates complete positive correlation, i.e. two variables completely change in the same direction in value; -1 indicates complete negative correlation, i.e. two variables completely change in opposite directions in value; 0 indicates no linear correlation between two variables. Pearson Correlation Coefficient is suitable for measuring the correlation between continuous variables, but may not be applicable in the case of nonlinear relationship or abnormal values. It has wide application in data analysis, scientific research, finance, etc., helping to identify the relationship between variables, thus conducting more in-depth analysis and prediction.

[0034] The information collected in the embodiments of the present application is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of related data comply with relevant laws, regulations and standards of countries and regions, necessary security measures are taken, do not violate public order and good customs, and provide corresponding operation portal for user to choose authorization or refusal.

[0035] Embodiment 1

[0036] According to the embodiments of the present application, a user demand processing method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from here.

[0037] Figure 1 is a flowchart of a user appeal processing method according to an embodiment of the present application, as shown in the figure, the method comprises the following steps: Figure 1

[0038] Step S102, in response to the user appeal information of the target user, extracting a plurality of information keywords in the user appeal information, and determining a target appeal level of the target user according to the plurality of information keywords.

[0039] Step S104, determining a target appeal information feature library and a target appeal intention library corresponding to the target appeal level, wherein the target appeal information feature library stores a plurality of appeal information features, and the target appeal intention library stores a plurality of appeal intentions corresponding to different appeal information features.

[0040] Step S106, obtaining a target user portrait of the target user, and extracting a plurality of target user information features from the target user portrait.

[0041] Step S108, determining a plurality of candidate appeal information features matched with the plurality of information keywords from the target appeal information feature library, and determining a target appeal information feature matched with the plurality of target user information features from the plurality of candidate appeal information features.

[0042] Step S110, determining a target appeal intention corresponding to the target appeal information feature from the target appeal intention library, and generating appeal feedback information according to the target appeal intention, and feeding back the appeal feedback information to the target user.

[0043] The steps of the user appeal processing method will be described in detail below in combination with the specific implementation process.

[0044] As an optional implementation, Figure 2 shows a flowchart of an optional target appeal level determination method. Specifically, the target appeal level of the target user can be determined according to the plurality of information keywords by the following way:

[0045] Step S1, performing part-of-speech tagging on the plurality of information keywords respectively to obtain part-of-speech tagging results.

[0046] Suppose we receive the target user appeal information as: "I am very dissatisfied with your service, the problem handling efficiency is too low, it is simply unacceptable.", for this appeal information, we can determine the target appeal level of the target user by the following way.

[0047] ​First, the information keywords in this appeal information are tagged with parts of speech using natural language processing tools. For example, "service" is tagged as a noun (assigned a value of 1), "unsatisfied" and "too low efficiency" are tagged as adjectives (assigned a value of 3), and "handle the problem" is tagged as a verb phrase (assigned a value of 2). Part-of-speech tagging helps understand the grammatical function and semantic role of keywords in a sentence.

[0048] Step S2, respectively, the sentiment analysis of a plurality of information keywords, get sentiment analysis results.

[0049] Next, sentiment analysis is performed on these information keywords. The analysis shows that "unsatisfied" and "unacceptable" have strong negative emotional tendencies, and the emotional intensity value may be -0.9; "too low efficiency in handling problems" also has negative emotions, and the intensity value may be -0.6. Use sentiment analysis algorithms (such as dictionary-based sentiment analysis, machine learning-based sentiment analysis models) to classify information keywords by sentiment, determine their sentiment as positive, negative, or neutral, and for keywords with sentiment, calculate their emotional intensity value, which is between -1 and 1, -1 indicating extreme negative emotion, 1 indicating extreme positive emotion, and 0 indicating neutral emotion.

[0050] Step S3, according to the part-of-speech tagging results and the sentiment analysis results, construct the keyword feature vector corresponding to the user appeal information.

[0051] Then, the part-of-speech tagging results and the sentiment analysis results are integrated to construct the keyword feature vector. For example, the keyword feature value corresponding to "service" is: 1 x (-0.1) = (-0.1), because usually the word "service" itself does not have strong emotional color, this is just an example; the keyword feature value corresponding to "unsatisfied" is: 3 x (-0.9) = (-2.7); the keyword feature value corresponding to "handle the problem" is: 2 x (-0.3) = (-0.6); the keyword feature value of "too low efficiency" is: 3 x (-0.6) = (-1.8).

[0052] Step S4, calculate the first similarity between the keyword feature vector and the preset feature vector.

[0053] Subsequently, based on the constructed keyword feature vector, we can calculate the first similarity between the keyword feature vector and the preset feature vector. Suppose there is a preset feature vector [RE1 RE2 RE i ……RE n ] corresponding to the keyword feature values in each position of the above keyword feature vector [RY1 RY2 RY3 RY4], the first similarity can be calculated as follows:

[0054]

[0055] In the formula, RP is the first similarity, i ∈ [1, n]. In combination with the above example, since there are 4 keyword feature values in the keyword feature vector, n can be at most 4. Assuming that the calculation result of RP is 0.75, this indicates that the keyword features in the user appeal information have a high similarity with the preset negative sentiment feature vector.

[0056] In step S5, the target appeal level corresponding to the first similarity is determined from a preset appeal level mapping table, wherein the mapping relationship between different similarities and different appeal levels is stored in the appeal level mapping table.

[0057] Finally, the corresponding target appeal level is found in the preset appeal level mapping table according to the first similarity RP value. In this embodiment, since the RP value is less than the preset degree value (for example, 0.8, which is only an example and does not constitute a specific limitation), according to the mapping table, it is determined that the target appeal level corresponding to the user appeal information is “high”, indicating that this is a user appeal with high negative sentiment intensity, which needs to be immediately noticed and prioritized by the enterprise.

[0058] Through the above process, we can see that the application can deeply understand the keyword semantics in the user appeal and quantify the emotional tendency through the part-of-speech tagging and sentiment analysis of the information keywords; by constructing a keyword feature vector and comparing it with a preset feature vector, the first similarity is quickly calculated, and the target appeal level is further determined, thereby realizing the automatic and efficient analysis of the user appeal, which helps the enterprise to accurately locate the high-priority user feedback, take timely measures, and improve customer satisfaction and enterprise service efficiency.

[0059] Suppose we have determined the target user's appeal information to be of high appeal degree according to the previous steps, which means that the problem or dissatisfaction raised by the user needs to be solved quickly and professionally. Next, we will determine the target appeal information feature library and the target appeal intent library corresponding to this level.

[0060] As an optional implementation, the target appeal information feature library and the target appeal intent library corresponding to the target appeal level can be determined by the following method: determining the target appeal information feature library corresponding to the target appeal level from a preset appeal information feature library mapping table, wherein the mapping relationship between different appeal levels and different appeal information feature libraries is stored in the appeal information feature library mapping table; determining the target appeal intent library corresponding to the target appeal level from a preset appeal intent library mapping table, wherein the mapping relationship between different appeal levels and different appeal intent libraries is stored in the appeal intent library mapping table.

[0061] The target appeal information feature library corresponding to the high appeal level is found from the preset appeal information feature library mapping table. The appeal information feature library mapping table can be established based on historical data and expert experience, and it contains the mapping relationship between different appeal levels and corresponding appeal information features. For example, the high appeal level can be mapped to a feature library containing multiple appeal information features such as bad attitude, product major failure, and low processing efficiency.

[0062] The target appeal intention library corresponding to the high appeal level is found from the preset appeal intention library mapping table. The appeal intention library mapping table can also be constructed based on historical data and expert experience, and it associates different appeal levels with a series of appeal intentions such as requesting a refund, seeking technical support, complaining about service, and inquiring about product details. The high appeal level can be mapped to an intention library containing multiple appeal intentions such as requesting urgent resolution of the problem, requiring service compensation, and seeking management intervention.

[0063] Once the target appeal information feature library and the target appeal intention library matching the high appeal level are determined, we can further analyze the user's specific appeal. For example, the user appeal contains the keywords "very dissatisfied", "low processing efficiency", and "unacceptable", which correspond to the high appeal level. The high appeal level can be mapped to information feature library A, and according to the keywords, it can be mapped to the appeal features in information feature library A that match these keywords, such as "processing time exceeded" and "service attitude problem"; the high appeal level can be mapped to intention appeal library B, and according to the keywords, it can be mapped to the appeal intentions in information feature library B that match these keywords, such as "requesting urgent resolution of the problem" and "requiring service compensation". This indicates that the user hopes to receive timely and effective remedial measures to compensate for the inconvenience they have experienced.

[0064] Finally, based on the identified specific appeal features and intentions, we can construct targeted response strategies. In practical applications, we can also dynamically adjust the mapping table to continuously optimize the matching strategy based on historical data and user feedback, further improving the accuracy and satisfaction of processing.

[0065] To further understand the user's appeal, we also integrate the information features of the user portrait to obtain more accurate appeal information features, achieve more in-depth personalized services, and more accurate problem processing.

[0066] As an optional implementation, Figure 3 An optional flowchart of a target user portrait acquisition method is shown. Specifically, the target user portrait of the target user can be acquired by the following method:

[0067] Step S1, obtain the user data of the target user, and pre-process the user data, wherein the pre-processing includes at least one of the following: data cleaning, standardization processing.

[0068] We can first obtain the target user's personal information data (such as age, gender, etc.), behavior data (website or APP usage frequency, click rate, etc.), transaction data (purchase records, order details, etc.), feedback data (previous complaints, suggestions, etc.), and external data (social media activity, interests, etc.) from the target user's historical data.

[0069] And pre-process the above data, including but not limited to data cleaning (remove invalid or duplicate data), standardization processing (such as uniform data format, scale, etc.), to ensure the quality and consistency of the data for subsequent analysis.

[0070] Step S2, use the pre-trained user classification model to analyze the pre-processed user data, extract the target user's target user features, and determine the target user's target user type.

[0071] Use the pre-trained user classification model to analyze the pre-processed user data, the model may be based on deep learning or machine learning algorithms, trained by historical data, can identify different types of users. For example, based on the data information of the target user that has been extracted, the model may analyze that the target user belongs to "high-value customers" and "service-sensitive customers", which indicates that the user not only has a large consumption, but also has strict requirements on service quality and efficiency.

[0072] As an optional implementation, multiple target user information features can be extracted from the target user portrait by the following method: extracting multi-dimensional user information features from the target user portrait, and standardizing the multi-dimensional user information features; clustering the standardized multi-dimensional user information features to obtain multiple clustering clusters; taking the user information features corresponding to the clustering center of each clustering cluster as the target user information features.

[0073] Assuming the user's complaint information is: "I am extremely dissatisfied with your service, the problem handling efficiency is too low, it is simply unacceptable." We first construct the target user portrait, assuming the user is a 45-year-old male who frequently uses telecommunications services. This portrait may be based on the user's historical behavior data, personal profile, and feedback information, etc. From this portrait, we extract multiple dimensions of user information features, including but not limited to the user's age, gender, frequency of using telecommunications services (service usage frequency), frequency of past complaints (historical complaint frequency), average waiting time for processing (average waiting time), and satisfaction rating for service (satisfaction level). These features are all based on user portraits, quantitative descriptions of user past behavior and preferences.

[0074] After extracting the features, standardization processing is performed to ensure that all features are compared and analyzed on the same scale and distribution. For example, the age feature may be directly in "years", while the satisfaction rating may be in "stars" (1 to 5 stars), and standardization processing can convert the age feature into "age group" (such as middle-aged population), and convert the satisfaction rating into "satisfaction level" (such as low, medium, high). This step ensures that different types of features can be fairly considered in subsequent clustering analysis, avoiding analysis bias caused by dimension or distribution.

[0075] The standardized user information features are converted into feature vectors. In the above example, we may obtain a feature vector containing age group, gender, service usage frequency, historical complaint frequency, average waiting time, and satisfaction level, each vector representing a set of feature values of the user in a specific dimension.

[0076] Relevant clustering algorithms can be used to cluster the feature vectors to form multiple clusters, and the number of clusters and algorithm selection depend on the specific situation of the data and the analysis goal. For example, we can determine the optimal number of clusters by calculating the silhouette coefficient or applying the elbow rule.

[0077] Each cluster represents a group of users with similar behavior patterns and satisfaction levels. We can observe the distribution of features within each cluster through data visualization tools to understand the common characteristics of users, such as a cluster may contain those who frequently complain and are particularly sensitive to efficiency.

[0078] For each cluster, we calculate its cluster center, which is the average value of the information feature vectors of all users in the cluster. The cluster center is a "virtual user" that integrates the characteristics of all users in the cluster, and its characteristic value can summarize the behavior and preferences of users in the cluster. The cluster center characteristics of each cluster are determined as target user information features. These features are targeted at a specific user group and reflect the core attributes and needs of the group. For example, if users in a cluster generally have a high complaint frequency and a high sensitivity to service efficiency, then "high complaint frequency" and "efficiency sensitivity" will be determined as target user information features. The target user information features can be recorded as a feature vector, assuming: [TY1 TY2 TY3TY4], which is used to subsequently adjust the standard user profile to obtain the target user profile of the target user.

[0079] Step S3: determining the target demand scenario based on the user demand information.

[0080] Based on the user’s demand information “I am very dissatisfied with your service. The efficiency of handling problems is too low and it is simply unacceptable.” It can be identified that the target demand scenario is the “complaint feedback scenario”. This scenario identifier reflects the user’s current dissatisfaction with the service and concern about efficiency issues.

[0081] Step S4: Determine the target standard user portrait corresponding to the target user type and target demand scenario from the preset user portrait library, wherein the user portrait library stores the mapping relationship between different user types and different demand scenarios and different standard user portraits, and each standard user portrait includes the standard user characteristics of a type of user in a demand scenario and the corresponding standard portrait parameters.

[0082] Based on the target user type (high-value customers and service-sensitive customers) and the target demand scenario (complaint feedback scenario), the best-matching standard user profile is found from the pre-set user profile library. The user profile library stores standard profiles for different user types in different demand scenarios. Each standard profile includes the standard characteristics and parameters of that user type in that scenario, such as "urgent need for efficient service" and "sensitivity to speed in problem resolution."

[0083] Step S5: Adjust the target standard user portrait according to the target user characteristics to obtain a target user portrait of the target user.

[0084] Based on the target user's specific characteristics and demands, adjust the standard user profile to better suit the target user's situation. For example, if the target user has repeatedly complained about service efficiency in the past, the "problem resolution speed" parameter can be increased in the standard profile parameters to ensure that the user's demands are prioritized and responded to promptly in subsequent services.

[0085] As an optional implementation, the target user portrait of the target user is obtained by adjusting the target standard user portrait according to the target user characteristics, which can be realized by the following method: determining the target standard user characteristics in the target standard user portrait and the corresponding target standard portrait parameters; calculating the second similarity between the target user characteristics and the target standard user characteristics, and adjusting the target standard portrait parameters with the second similarity as the adjustment coefficient; generating the target user portrait of the target user according to the target standard user characteristics and the adjusted target standard portrait parameters.

[0086] The specific adjustment can include the following steps in the above step S5:

[0087] Step S51, determine the user characteristic index.

[0088] For example, the target user characteristics are frequent complaints about inefficiency, the standard user characteristics are dissatisfaction with inefficient services, and the user characteristic index TP is calculated as the cosine similarity of the feature values, which reflects the degree of fit between the target user and the standard portrait in a specific feature. Assuming that there are preset values in the standard user characteristic vector [TE1 TE2 TE j ……TE m ] corresponding to the target user characteristic values in each position of the target user characteristic vector [TY1 TY2 TY3 TY4], the user characteristic index TP can be calculated by the following method:

[0089]

[0090] In the formula, TP is the user characteristic index, and j ∈ [1, m]. The value of m can be determined according to the number of user characteristic values in the target user characteristic vector of the user. According to the above example, m can be taken as 4 at most.

[0091] Step S52, adjust the standard portrait parameters.

[0092] According to the user characteristic index TP and the weight W of the standard portrait parameters, the new weight W ′ of the adjusted parameters can better reflect the personalized needs of the target user.

[0093] The adjustment of the user portrait is based on the comparison of the user characteristics and the standard portrait characteristics, and the portrait parameters are fine-tuned by calculating the similarity. For example, if the user often uses the service at night, the parameter of night activity will be increased. This method ensures the dynamic and personalized nature of the user portrait, and can more accurately reflect the real needs and preferences of the user, solving the problem of static and generalization of the user portrait.

[0094] Through these adjustments, the resulting target user profile will include the target user type (high-value customers and service-sensitive customers), the target demand scenario (complaint feedback scenario), and the adjusted personalized parameters. By accurately matching user profiles and demand scenarios, more personalized and effective feedback information can be generated, effectively resolving the mismatch between feedback information and user needs.

[0095] As an optional implementation, determining multiple candidate appeal information features that match multiple information keywords from a target appeal information feature library can be achieved in the following manner: for each information keyword, determining multiple target information fuzzy words whose semantic similarity with the information keyword is greater than a preset similarity threshold from a preset information fuzzy word library, wherein the information fuzzy word library stores multiple information fuzzy words; determining the appeal information feature corresponding to each target information fuzzy word as a candidate appeal information feature from the target appeal information feature library, wherein the target appeal information feature library stores a mapping relationship between different information fuzzy words and different appeal information features.

[0096] Assume that we have identified obvious information keywords such as "unsatisfactory service" and "inefficient problem handling" based on the appeal information. For these keywords, we can further perform semantic analysis to identify their corresponding information-ambiguous words. For example, for "unsatisfactory service", possible information-ambiguous words include "poor service", "inconsiderate service", "improper handling", etc.; for "inefficient problem handling", possible information-ambiguous words include "slow response", "delayed", "long waiting time", etc.

[0097] The information fuzzy word library contains a collection of words with similar meanings to the original information keywords. We set a semantic similarity threshold, such as 0.6, to filter out information fuzzy words with high similarity. This threshold can be adjusted based on actual conditions to balance information coverage and accuracy.

[0098] Based on the preset similarity threshold, we can find out from the information fuzzy word library the words with semantic similarity higher than 0.6 to the information keywords "unsatisfied service" and "low efficiency in handling problems". For example, words such as "poor service" and "slow response" may be selected because they are close in meaning to the original keywords and can capture the user's potential semantic information, enriching the understanding of the appeal. From the target appeal information feature library, we determine the appeal information features corresponding to each target information fuzzy word. The appeal information feature library stores the mapping relationship between different information fuzzy words and appeal information features, including service features (service attitude, service efficiency, etc.), product features, price features, order features, user experience features, sales features, feedback and suggestion features, etc. For example, for the information fuzzy word "poor service", it may be mapped to "poor service quality" in the service feature. For each information fuzzy word, we find the corresponding appeal information features from the appeal information feature library, which constitute the candidate appeal information feature set.

[0099] Through the information fuzzy word library and the similarity threshold, the potential meaning and extended range of user appeal information can be identified, such as associating "unsatisfied service" with "poor service", "inadequate service", "improper handling", etc., to better understand user appeals. This method can generate more comprehensive and accurate appeal information features by deeply mining the semantic association of keywords, solving the problem of incomplete keyword recognition.

[0100] As an optional implementation, determining the target appeal information feature that matches the target user information feature from multiple candidate appeal information features can be achieved by the following method: for each candidate appeal information feature, calculate the degree of association between the candidate appeal information feature and each target user information feature, and take the average of the multiple degrees of association as the comprehensive degree of association between the candidate appeal information feature and the target user, wherein the degree of association includes one of the following: cosine similarity, Pearson correlation coefficient; determine the candidate appeal information feature with the maximum comprehensive degree of association with the target user as the target appeal information feature.

[0101] Suppose the key appeal information features are "low service efficiency", "long processing time", and "long waiting response time", and the target user information features are: age (middle-aged), occupation (IT engineer), historical complaint frequency (high), service satisfaction (low), usage frequency (high), demand for efficiency (high), etc.

[0102] For each key information feature, we calculate its association degree with each target user information feature. For example, for "low service efficiency", its association degree with the target user information feature "high demand for efficiency" may be very high, because IT engineers have high expectations for service efficiency; while its association degree with "age (middle-aged)" may be lower, because age has little direct relationship with service efficiency. We can use statistical methods such as cosine similarity or Pearson correlation coefficient to calculate. Using statistical methods to quantify the association between key information features and user portraits provides data support for the optimization of customer service strategies, and helps enterprises make more scientific and accurate service decisions.

[0103] For each key information feature, we calculate its association degree with each target user information feature. For example, for "low service efficiency", its association degree with the target user information feature "high demand for efficiency" may be very high, because IT engineers have high expectations for service efficiency; while its association degree with "age (middle-aged)" may be lower, because age has little direct relationship with service efficiency. We can use statistical methods such as cosine similarity or Pearson correlation coefficient to calculate. Using statistical methods to quantify the association between key information features and user portraits provides data support for the optimization of customer service strategies, and helps enterprises make more scientific and accurate service decisions.

[0104] After obtaining the comprehensive association degree of all key information features, we sort them and select the key information feature with the highest comprehensive association degree as the target information feature. This feature best represents the user's current most urgent needs and problems. This process fully utilizes the multi-dimensional information of user portraits and achieves intelligent and personalized service response through data-driven methods.

[0105] As an optional implementation, generating the feedback information according to the target information feature can be achieved by the following method: determining a target output template corresponding to the target information feature from a preset output template library, wherein the output template library stores mapping relationships between different information features and different output templates; determining a target output corpus corresponding to the target output template and the target user portrait from a preset output corpus library, wherein the output corpus library stores mapping relationships between different output templates, different user portraits, and different output corpora; and filling the target output corpus into the target output template to obtain the feedback information.

[0106] Assuming we've identified the user's intent as "requesting an expedited resolution," we can identify an output template corresponding to this intent from a pre-set output template library. This template might include key components such as the problem description, impact analysis, resolution plan, expected improvement timeline, and compensation plan. The output template library is built based on extensive historical data and expert knowledge, ensuring the structured and standardized output information, facilitating the rapid generation of high-quality feedback. The output template library can be built based on extensive historical data and expert knowledge, ensuring the structured and standardized output information, facilitating the rapid generation of high-quality feedback.

[0107] For example, our target user is a 45-year-old male who frequently uses telecommunications services and is sensitive to efficiency. Based on this user profile, we select corpora from our pre-defined output corpus that are suitable for middle-aged, efficiency-sensitive users. This output corpus contains different expressions, case studies, and solution examples for different user types, ensuring that the information output is both professional and tailored to the preferences of this specific user group. For example, for this user, we might use more formal, direct, and problem-solving-oriented corpora.

[0108] Finally, the determined target output template is combined with the target output corpus, specific information is filled in, and appeal feedback information is generated.

[0109] Through template-based processing, the system can quickly generate clearly structured and comprehensive feedback, significantly reducing response time and improving processing efficiency. By selecting appropriate output corpus based on user profiles, the system ensures that the output aligns with the user's language style and information preferences, improving user acceptance and satisfaction. The combination of standardized output templates and personalized corpus reduces ambiguity in information communication and ensures that users accurately understand the company's response and solutions.

[0110] After obtaining the demand feedback information, the demand feedback information can not only be pushed to the target users, but also directly pushed to user groups with similar needs.

[0111] As an optional implementation, after feeding back the appeal feedback information to the target user, the method further includes: determining a target user group corresponding to the target user type; and pushing the appeal feedback information to all users in the target user group.

[0112] We can build a user classification model based on historical data to determine the type of target users. Historical data includes personal information, behavior data, transaction records, feedback information, etc. of users, through which a multi-dimensional portrait of users is constructed, and a machine learning model is trained for user classification. The classification criteria may include the user's age stage, occupation category, usage habit, historical feedback type, etc. to identify user groups with similar behavior patterns and demand characteristics.

[0113] For target users, we can extract key portrait parameters from the user portrait, which may include the user's occupation, age, usage frequency, historical feedback type, etc. to further subdivide user groups. For example, if the target user is an IT engineer who frequently uses telecommunications services, the key portrait parameters may include "high-frequency user", "IT industry", "high complaint rate", etc.

[0114] By inputting the key portrait parameters of the target user into the user classification model, the model identifies other users with similar characteristics to the target user, forming a target user group. For example, the model may identify other high-frequency users, IT industry practitioners, or user groups with similar complaint histories who may also face similar service efficiency problems. Even if these users have not made explicit demands, demand feedback information can be pushed to all users in the target user group corresponding to the target user type. In this way, enterprises can proactively take measures to address potential problems before they spread, thereby improving user satisfaction and reducing the number of complaints.

[0115] Through the above steps, key information is extracted from user complaint information to quickly identify the core of user complaints, avoiding the interpretation bias that may exist in manual customer service and improving the speed and accuracy of problem positioning; by establishing a complaint information feature library and a complaint intent library associated with the complaint level, the system can call the appropriate information features and intents based on the specific content of the user complaint, forming a structured processing flow. This not only reduces the task burden of manual customer service, but also makes the system's response more standardized and precise, avoiding inconsistencies in service due to individual differences in customer service. Combined with user portrait information, personalized responses that meet user needs are made; through keyword matching with the information feature library, the system can accurately identify the essence of user complaints and further refine them based on user information features, improving the relevance of problem handling; after determining the target complaint intent, the system can automatically generate feedback information based on pre-set templates or algorithms, not only improving the speed of information transmission, but also ensuring the professionalism and consistency of feedback information, thereby solving the technical problems of traditional manual customer service in handling user complaints, such as low efficiency, poor accuracy, and difficulty in meeting user needs.

[0116] Embodiment 2

[0117] According to the embodiments of the present application, a user appeal processing device for implementing the user appeal processing method in embodiment 1 is also provided, as shown in the figure, the user appeal processing device at least includes: a level determination module 41, a library selection module 42, a user feature determination module 43, an appeal feature determination module 44, and an appeal feedback module 45, wherein: Figure 4

[0118] The level determination module 41 can extract a plurality of information keywords in the user appeal information in response to the user appeal information of the target user, and determine the target appeal level of the target user according to the plurality of information keywords.

[0119] The library selection module 42 can determine a target appeal information feature library and a target appeal intent library corresponding to the target appeal level, wherein the target appeal information feature library stores a plurality of appeal information features, and the target appeal intent library stores a plurality of appeal intents corresponding to different appeal information features.

[0120] The user feature determination module 43 can obtain a target user portrait of the target user, and extract a plurality of target user information features from the target user portrait.

[0121] The appeal feature determination module 44 can determine a plurality of candidate appeal information features matching the plurality of information keywords from the target appeal information feature library, and determine a target appeal information feature matching the plurality of target user information features from the plurality of candidate appeal information features.

[0122] The appeal feedback module 45 can determine a target appeal intent corresponding to the target appeal information feature from the target appeal intent library, and generate appeal feedback information according to the target appeal intent, and feed back the appeal feedback information to the target user.

[0123] The functions of each module of the user appeal processing device will be described in detail below in combination with the specific implementation process.

[0124] As an optional implementation manner, the level determination module determines the target appeal level of the target user according to the plurality of information keywords, which can be implemented by the following manner: performing part-of-speech tagging on the plurality of information keywords respectively to obtain part-of-speech tagging results; performing sentiment analysis on the plurality of information keywords respectively to obtain sentiment analysis results; constructing a keyword feature vector corresponding to the user appeal information according to the part-of-speech tagging results and the sentiment analysis results; calculating a first similarity between the keyword feature vector and a preset feature vector; determining a target appeal level corresponding to the first similarity from a preset appeal level mapping table, wherein the appeal level mapping table stores a mapping relationship between different similarities and different appeal levels.

[0125] ​As an optional implementation, the library selection module determines the target appeal information feature library and the target appeal intent library corresponding to the target appeal level, which can be achieved by the following manner: determining the target appeal information feature library corresponding to the target appeal level from a preset appeal information feature library mapping table, wherein the appeal information feature library mapping table stores mapping relationships between different appeal levels and different appeal information feature libraries; determining the target appeal intent library corresponding to the target appeal level from a preset appeal intent library mapping table, wherein the appeal intent library mapping table stores mapping relationships between different appeal levels and different appeal intent libraries.

[0126] As an optional implementation, the user feature determination module obtains the target user portrait of the target user, which can be achieved by the following manner: obtaining user data of the target user, and pre-processing the user data, wherein the pre-processing includes at least one of the following: data cleaning, standardization processing; analyzing the pre-processed user data by using a pre-trained user classification model, extracting target user features corresponding to the target user, and determining a target user type of the target user; determining a target appeal scene according to the user appeal information; determining a target standard user portrait corresponding to the target user type and the target appeal scene from a preset user portrait library, wherein the user portrait library stores mapping relationships between different user types, different appeal scenes and different standard user portraits, and each standard user portrait includes standard user features and corresponding standard portrait parameters of a type of user in an appeal scene; adjusting the target standard user portrait according to the target user features to obtain the target user portrait of the target user.

[0127] As an optional implementation, the user feature determination module adjusts the target standard user portrait according to the target user features to obtain the target user portrait of the target user, which can be achieved by the following manner: determining target standard user features and corresponding target standard portrait parameters in the target standard user portrait; calculating a second similarity between the target user features and the target standard user features, and adjusting the target standard portrait parameters by taking the second similarity as an adjustment coefficient; generating the target user portrait of the target user according to the target standard user features and the adjusted target standard portrait parameters.

[0128] As an optional implementation, the user feature determination module extracts a plurality of target user information features from the target user portrait, which can be achieved by the following manner: extracting multi-dimensional user information features from the target user portrait, and performing standardization processing on the multi-dimensional user information features; clustering the multi-dimensional user information features after the standardization processing to obtain a plurality of clustering clusters; taking user information features corresponding to a clustering center of each clustering cluster as the target user information features.

[0129] As an optional implementation, the appeal feature determination module determines the multiple candidate appeal information features matched with the multiple information keywords from the target appeal information feature library, which can be implemented in the following manner: for each information keyword, multiple target information fuzzy words with a semantic similarity greater than a preset similarity threshold to the information keyword are determined from a preset information fuzzy word library, wherein the information fuzzy word library stores multiple information fuzzy words; an appeal information feature corresponding to each target information fuzzy word is determined from the target appeal information feature library as a candidate appeal information feature, wherein the target appeal information feature library stores a mapping relationship between different information fuzzy words and different appeal information features.

[0130] As an optional implementation, the appeal feature determination module determines the target appeal information feature matched with the multiple target user information features from the multiple candidate appeal information features, which can be implemented in the following manner: for each candidate appeal information feature, the correlation degree between the candidate appeal information feature and each target user information feature is calculated respectively, and the average value of the multiple correlation degrees is taken as the comprehensive correlation degree between the candidate appeal information feature and the target user, wherein the correlation degree includes one of the following: cosine similarity, Pearson correlation coefficient; the candidate appeal information feature with the maximum comprehensive correlation degree with the target user is determined as the target appeal information feature.

[0131] As an optional implementation, the appeal feedback module generates the appeal feedback information according to the target appeal intention, which can be implemented in the following manner: a target output template corresponding to the target appeal intention is determined from a preset output template library, wherein the output template library stores a mapping relationship between different appeal intentions and different output templates; a target output corpus corresponding to the target output template and the target user portrait is determined from a preset output corpus library, wherein the output corpus library stores a mapping relationship between different output templates and different user portraits and different output corpora; the target output corpus is filled into the target output template to obtain the appeal feedback information.

[0132] As an optional implementation, after the appeal feedback information is fed back to the target user, the appeal feedback module can further determine a target user group corresponding to the target user type; the appeal feedback information is pushed to all users in the target user group.

[0133] It should be noted that each module in the user appeal processing apparatus in the embodiments of the present application corresponds to each implementation step of the user appeal processing method in Embodiment 1, and since Embodiment 1 has been described in detail, the details not embodied in the present embodiment can be referred to Embodiment 1, and will not be described in more detail here.

[0134] Embodiment 3

[0135] According to an embodiment of the present application, a computer program product is provided, which comprises a computer program. The computer program, when executed by a processor, implements the user appeal processing method in embodiment 1.

[0136] According to an embodiment of the present application, a non-volatile storage medium is provided, which comprises a stored computer program. The device in which the non-volatile storage medium is located executes the user appeal processing method in embodiment 1 by running the computer program.

[0137] According to an embodiment of the present application, a processor is provided, which is used to run a computer program. The computer program, when executed, executes the user appeal processing method in embodiment 1.

[0138] According to an embodiment of the present application, an electronic device is provided, which comprises a memory and a processor. The memory stores a computer program, and the processor is configured to execute the user appeal processing method in embodiment 1 by the computer program.

[0139] Specifically, the computer program, when executed, implements the following steps: in response to the user appeal information of the target user, extracting a plurality of information keywords in the user appeal information, and determining a target appeal level of the target user according to the plurality of information keywords; determining a target appeal information feature library and a target appeal intent library corresponding to the target appeal level, wherein the target appeal information feature library stores a plurality of appeal information features, and the target appeal intent library stores a plurality of appeal intents corresponding to different appeal information features; obtaining a target user portrait of the target user, and extracting a plurality of target user information features from the target user portrait; determining a plurality of candidate appeal information features matched with the plurality of information keywords from the target appeal information feature library, and determining a target appeal information feature matched with the plurality of target user information features from the plurality of candidate appeal information features; determining a target appeal intent corresponding to the target appeal information feature from the target appeal intent library, and generating appeal feedback information according to the target appeal intent, and feeding back the appeal feedback information to the target user.

[0140] As an optional implementation, the electronic device can exist in the form of a mobile terminal, a computer terminal or a similar computing device. Figure 5 A hardware structure block diagram of an electronic device for implementing a user appeal processing method is shown. As shown in FIG. 1, the electronic device comprises a processor 11 and a memory 12. Figure 5As shown, the electronic device 50 can include one or more processors 502 (the processor 502 can include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 504 for storing data, and a transmission device 506 for communication functions. In addition, it can also include a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which can be included as one of the ports in the BUS bus), a network interface, a power supply and / or a camera. Those skilled in the art can understand that Figure 5 The structure shown is only schematic, which does not limit the structure of the above-mentioned electronic device. For example, the electronic device 50 can also include more or fewer components than those shown in Figure 5 or have a different configuration than that shown in Figure 5 .

[0141] It should be noted that the one or more processors 502 and / or other data processing circuits described above can be referred to herein as "data processing circuits" in general. The data processing circuit can be embodied in whole or in part as software, hardware, firmware or any combination thereof. In addition, the data processing circuit can be a single independent processing module, or all or part of any one of the other elements combined into the electronic device 50. As referred to in the embodiments of the present application, the data processing circuit serves as a processor to control (for example, selection of a variable resistance terminal path connected to an interface).

[0142] The memory 504 can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to the user complaint processing method in the embodiments of the present application. The processor 502 executes various functional applications and data processing by running the software programs and modules stored in the memory 504, that is, implements the vulnerability detection method of the application program described above. The memory 504 can include a high-speed random access memory, and can also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 504 can further include a memory remotely disposed with respect to the processor 502, which can be connected to the electronic device 50 through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network and a combination thereof.

[0143] The transmission device 506 is configured to receive or send data via a network. The network can include a wireless network provided by a communication provider of the electronic device 50. In one example, the transmission device 506 includes a network interface controller (NIC) that can connect to other network devices through a base station to communicate with the Internet. In one example, the transmission device 506 can be a radio frequency (RF) module that is configured to communicate with the Internet wirelessly.

[0144] The display can be a liquid crystal display (LCD) that is touch screen type, for example, which can enable a user to interact with a user interface of the electronic device 50.

[0145] The above-mentioned embodiment numbers are only for description, and do not represent the advantages or disadvantages of the embodiments.

[0146] In the above-mentioned embodiments of the present application, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0147] In the several embodiments provided in the present application, it should be understood that the disclosed technology can be implemented in other ways. Of course, the embodiment described above is only a schematic, for example, the division of units can be a logical function division, and actual implementation can be another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, unit or module, and can be electrical or other forms.

[0148] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple units. Part or all of the units can be selected according to actual needs to achieve the purpose of the present embodiment.

[0149] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0150] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned storage medium includes: a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0151] The above is only the preferred embodiment of the present application, and it should be pointed out that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should be considered as the protection scope of the present application.

Claims

1. A method for handling user demands, characterized in that: include: In response to user demand information of a target user, extracting a plurality of information keywords from the user demand information, and determining a target demand level of the target user based on the plurality of information keywords; Determine a target appeal information feature library and a target appeal intention library corresponding to the target appeal level, wherein the target appeal information feature library stores a plurality of appeal information features, and the target appeal intention library stores a plurality of appeal intentions corresponding to different appeal information features; Obtaining a target user portrait of the target user, and extracting multiple target user information features from the target user portrait; Determine a plurality of candidate demand information features that match the plurality of information keywords from the target demand information feature library, and determine a target demand information feature that matches the plurality of target user information features from the plurality of candidate demand information features; The target appeal intention corresponding to the target appeal information feature is determined from the target appeal intention library, appeal feedback information is generated according to the target appeal intention, and the appeal feedback information is fed back to the target user.

2. The method according to claim 1, characterized in that Determining the target demand level of the target user based on the plurality of information keywords includes: Performing part-of-speech tagging on the plurality of information keywords respectively to obtain part-of-speech tagging results; Performing sentiment analysis on the plurality of information keywords respectively to obtain sentiment analysis results; Constructing a keyword feature vector corresponding to the user demand information based on the part-of-speech tagging result and the sentiment analysis result; Calculating a first similarity between the keyword feature vector and a preset feature vector; A target demand level corresponding to the first similarity is determined from a preset demand level mapping table, wherein the demand level mapping table stores mapping relationships between different similarities and different demand levels.

3. The method according to claim 1, characterized in that Determining a target appeal information feature library and a target appeal intention library corresponding to the target appeal level includes: Determining a target demand information feature library corresponding to the target demand level from a preset demand information feature library mapping table, wherein the demand information feature library mapping table stores mapping relationships between different demand levels and different demand information feature libraries; A target demand intention library corresponding to the target demand level is determined from a preset demand intention library mapping table, wherein the demand intention library mapping table stores mapping relationships between different demand levels and different demand intention libraries.

4. The method according to claim 1, wherein Obtaining a target user profile of the target user includes: Acquire user data of the target user and preprocess the user data, wherein the preprocessing includes at least one of the following: data cleaning and standardization; Analyze the pre-processed user data using a pre-trained user classification model, extract target user features corresponding to the target user, and determine the target user type of the target user; Determining a target demand scenario based on the user demand information; Determine a target standard user profile corresponding to the target user type and the target demand scenario from a preset user profile library, wherein the user profile library stores mapping relationships between different user types and different demand scenarios and different standard user profiles, and each standard user profile includes standard user features of a type of user in a demand scenario and corresponding standard profile parameters; The target standard user portrait is adjusted according to the target user characteristics to obtain a target user portrait of the target user.

5. The method according to claim 4, characterized in that Adjusting the target standard user profile according to the target user characteristics to obtain a target user profile of the target user includes: Determining target standard user characteristics and corresponding target standard profile parameters in the target standard user profile; Calculating a second similarity between the target user feature and the target standard user feature, and adjusting the target standard portrait parameters using the second similarity as an adjustment coefficient; A target user portrait of the target user is generated based on the target standard user characteristics and the adjusted target standard portrait parameters.

6. The method according to claim 4, characterized in that Extracting multiple target user information features from the target user portrait, including: Extracting multi-dimensional user information features from the target user portrait, and performing standardization processing on the multi-dimensional user information features; Clustering the standardized multi-dimensional user information features to obtain multiple clusters; The user information feature corresponding to the cluster center of each cluster is used as the target user information feature.

7. The method according to claim 1, characterized in that Determining a plurality of candidate appeal information features that match a plurality of information keywords from the target appeal information feature library includes: For each information keyword, determining a plurality of target information fuzzy words having a semantic similarity with the information keyword greater than a preset similarity threshold from a preset information fuzzy word library, wherein the information fuzzy word library stores a plurality of information fuzzy words; Determine the demand information feature corresponding to each of the target information fuzzy words from the target demand information feature library as the candidate demand information feature, wherein the target demand information feature library stores mapping relationships between different information fuzzy words and different demand information features.

8. The method according to claim 1, characterized in that Determining a target demand information feature that matches the multiple target user information features from the multiple candidate demand information features includes: For each candidate demand information feature, respectively calculate the correlation degree between the candidate demand information feature and each target user information feature, and use the average of multiple correlation degrees as the comprehensive correlation degree between the candidate demand information feature and the target user, wherein the correlation degree includes one of the following: cosine similarity, Pearson correlation coefficient; The candidate demand information feature having the greatest comprehensive correlation with the target user is determined as the target demand information feature.

9. The method according to claim 1, characterized in that Generate appeal feedback information based on the target appeal intention, including: Determining a target output template corresponding to the target appeal intent from a preset output template library, wherein the output template library stores mapping relationships between different appeal intents and different output templates; Determining a target output corpus corresponding to the target output template and the target user profile from a preset output corpus, wherein the output corpus stores mapping relationships between different output templates and different user profiles and different output corpora; The target output corpus is filled into the target output template to obtain the demand feedback information.

10. The method according to claim 4, characterized in that After feeding back the appeal feedback information to the target user, the method further includes: Determine a target user group corresponding to the target user type; The appeal feedback information is pushed to all users in the target user group.

11. A user appeal processing device, characterized in that: include: a level determination module, configured to extract a plurality of information keywords from the user demand information in response to the user demand information of the target user, and determine the target demand level of the target user based on the plurality of information keywords; a library selection module, configured to determine a target appeal information feature library and a target appeal intention library corresponding to the target appeal level, wherein the target appeal information feature library stores a plurality of appeal information features, and the target appeal intention library stores a plurality of appeal intentions corresponding to different appeal information features; A user feature determination module is used to obtain a target user portrait of the target user and extract multiple target user information features from the target user portrait; a demand feature determination module, configured to determine a plurality of candidate demand information features matching the plurality of information keywords from the target demand information feature library, and determine a target demand information feature matching the plurality of target user information features from the plurality of candidate demand information features; The appeal feedback module is used to determine the target appeal intention corresponding to the target appeal information feature from the target appeal intention library, generate appeal feedback information according to the target appeal intention, and feed back the appeal feedback information to the target user.

12. A computer program product, characterized in that include: A computer program, wherein when the computer program is executed by a processor, the method for processing user demands according to any one of claims 1 to 10 is implemented.

13. An electronic device, characterized in that: include: A memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the user demand processing method according to any one of claims 1 to 10 through the computer program.

Images