Customer service request data processing method, device, electronic device and storage medium
Through in-depth analysis of massive customer service data, natural language processing technology is used to extract event elements and generate abstracts, and trend analysis and early warning are carried out, which solves the problem of inefficiency in the existing technology, and realizes efficient and accurate data support and services, improving The timeliness and accuracy of corporate decision-making.
Patent Information
- Application Number
- CN202411117264.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-15
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-08-15
AI Technical Summary
The existing customer service demand situation intelligent perception and early warning system rely on structured data query statistics and visualization technology, which makes it difficult to achieve in-depth and accurate analysis of massive customer service data, and is inefficient, making it impossible to provide efficient and accurate data support and services.
By obtaining customer service appeal data, extracting appeal event titles and contents, generating event element information and summary information, conducting event trend analysis, generating early warning information, and pushing it to related terminals, using natural language processing technology for in-depth analysis, and achieving massive data Accurate understanding and real-time early warning.
It realizes comprehensive perception, accurate analysis and real-time early warning of customer service demands, provides efficient and accurate decision-making support for enterprise management, and improves enterprise competitiveness.
Smart Images

Figure CN118643828B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of natural language processing, and specifically to a method, device, electronic device and storage medium for processing customer service demand data. Background Art
[0002] With the increase in corporate service demand and the enhancement of customer participation awareness, customer service demands are becoming an increasingly important part of corporate management. Customer service demands not only reflect customers' expectations for service quality, but are also an important channel for companies to understand customer needs and optimize services. However, faced with massive amounts of customer service demand data, how to effectively perceive, analyze and warn of potential service problems has become a major challenge for corporate managers.
[0003] At present, the intelligent perception and early warning system of customer service demand situation mainly adopts structured data query statistics and visualization technology, which can realize data collection, storage, cleaning and preprocessing, generate statistical query results, and help enterprise managers have a macro understanding of customer service demands. However, the existing methods have some limitations: data analysis requires a lot of manual intervention, which is not only time-consuming and labor-intensive, but also inefficient, and cannot conduct in-depth and accurate analysis of massive customer service data, and cannot intuitively reflect the analysis results, thus failing to provide efficient and accurate data support and services for enterprise managers, and the limitations are relatively large.
[0004] The preceding description is intended to provide general background information and does not necessarily constitute prior art. Summary of the invention
[0005] The embodiments of the present application provide a customer service demand data processing method, device, electronic device and storage medium, which can perform in-depth and accurate analysis of massive customer service data and provide efficient and accurate data support and services for enterprise managers.
[0006] In a first aspect, an embodiment of the present application provides a method for processing customer service demand data, including:
[0007] Acquire customer service demand data, and extract demand event titles and demand event contents from the customer service demand data;
[0008] Extracting event element information based on the appeal event title and the appeal event content, and generating event summary information based on the appeal event content;
[0009] Performing event trend analysis on the target event based on the event element information and the event summary information to obtain an event trend analysis result;
[0010] Based on the event trend analysis result, corresponding warning information is generated, and the warning information is pushed to the associated terminal corresponding to the bound associated object.
[0011] Optionally, in some embodiments of the present application, the method further includes:
[0012] Based on the event trend analysis results and the warning information, generate a corresponding analysis report;
[0013] Based on the customer service demand data and the event trend analysis results, a corresponding demand event map is generated.
[0014] Optionally, in some embodiments of the present application, the method further includes:
[0015] Performing event clustering processing based on the event element information to obtain repeated appeal events;
[0016] Matching the information of interest based on the content of the appeal event to obtain a matching result, and determining the event of interest based on the matching result;
[0017] Weak signal analysis is performed based on the content of the appeal event to determine the weak signal event.
[0018] Optionally, in some embodiments of the present application, performing event trend analysis on the target event based on the event element information and the event summary information to obtain an event trend analysis result includes:
[0019] Respectively obtaining event element information and event summary information corresponding to the repeated appeal event, the event of interest, and the weak signal event;
[0020] Based on the event element information and the event summary information, event trend analysis is performed on the repeated demand events, the events of interest and the weak signal events respectively to obtain event trend analysis results corresponding to the repeated demand events, the events of interest and the weak signal events.
[0021] Optionally, in some embodiments of the present application, the extracting event element information based on the appeal event title and the appeal event content includes:
[0022] Combining the appeal event title and the appeal event content to obtain a corresponding text combination;
[0023] Using a word segmenter to perform word segmentation processing on the text combination to obtain a first word segmentation result;
[0024] Perform word embedding on the word segmentation result through a pre-trained model as input data for the first model;
[0025] Encoding the first model input data based on an encoder, and extracting corresponding text context-related features based on the encoded first model input data;
[0026] Use a multi-layer perceptron to classify and label each segmentation to obtain the entity type corresponding to each segmentation;
[0027] Based on the text context-related features and the entity type, adjacent word segments of the same entity type are merged to form event element information, and the event element information includes the place where the event occurred, the subject involved in the lawsuit, the event content and the event handling result.
[0028] Optionally, in some embodiments of the present application, generating event summary information based on the appeal event content includes:
[0029] Using a word segmenter to perform word segmentation processing on the content of the appeal event to obtain a second word segmentation result;
[0030] Perform word embedding on the word segmentation result through a pre-trained model as input data for the second model;
[0031] Using a decoder to decode the second model input data, and extracting corresponding text context-related features based on the decoded second model input data;
[0032] Generate a word probability distribution result corresponding to the second model input data through linear transformation and normalized exponential function;
[0033] Calculate the cumulative probability corresponding to each candidate queue in the word probability distribution result by a beam search algorithm;
[0034] A plurality of candidate queues with the highest cumulative probabilities are selected as event summary information.
[0035] Optionally, in some embodiments of the present application, performing event trend analysis on the target event based on the event element information and the event summary information to obtain an event trend analysis result includes:
[0036] Determine target events that meet the analysis criteria through preset business rules;
[0037] Arranging the target events in chronological order, and numerically encoding the event element information and the event summary information corresponding to the target events to obtain sequence analysis input data;
[0038] The sequence analysis input data is subjected to feature analysis and event trend analysis through a temporal convolutional network model to obtain an event trend analysis result corresponding to the target event.
[0039] Optionally, in some embodiments of the present application, performing event clustering based on the event element information to obtain repeated appeal events includes:
[0040] Perform word segmentation and word embedding processing on the event element information to obtain third model input data;
[0041] Inputting the third model input data into the semantic embedding model, and outputting the corresponding event vector through the self-attention layer and feedforward neural network of the semantic embedding model;
[0042] Using the cosine similarity function to calculate the similarity between all the event vectors, and construct a similarity matrix;
[0043] Based on the density clustering algorithm and the similarity matrix, event clustering processing is performed on the target event to obtain repeated appeal events.
[0044] Optionally, in some embodiments of the present application, matching the information of interest based on the content of the appeal event to obtain a matching result, and determining the event of interest based on the matching result includes:
[0045] Build a vocabulary of interest based on domain knowledge and experience;
[0046] Based on the interest word library, a keyword matching method is used to search for keywords in the content of the appeal event as the first interest information;
[0047] Using a word segmenter to segment the content of the appeal event and remove stop words, and searching for keywords in the segmentation results after removing the stop words by keyword matching;
[0048] Performing feature encoding on the keyword through a semantic encoding model, and calculating the semantic similarity between the feature-encoded keyword and the vocabulary of interest through a cosine similarity function;
[0049] Determining second information of interest in the content of the appeal event based on the semantic similarity;
[0050] Based on the first information of interest and the second information of interest, it is determined whether the target event is an event of interest, and whether the target event contains information of interest and a hit keyword of interest.
[0051] In a second aspect, an embodiment of the present application provides a customer service demand data processing device, including:
[0052] A data acquisition module, used to acquire customer service demand data and extract demand event titles and demand event contents from the customer service demand data;
[0053] A data analysis module, configured to extract event element information based on the appeal event title and the appeal event content, and generate event summary information based on the appeal event content;
[0054] An event analysis module, used to perform event trend analysis on a target event based on the event element information and the event summary information to obtain an event trend analysis result;
[0055] The event warning module is used to generate corresponding warning information based on the event trend analysis result, and push the warning information to the associated terminal corresponding to the bound associated object.
[0056] Correspondingly, the present application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program according to the steps of the method described in the first aspect.
[0057] The present application also provides a readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the computer program implements the method described in the first aspect.
[0058] The present application provides a customer service demand data processing method, device, electronic device and storage medium, the customer service demand data processing method comprising: obtaining customer service demand data, and extracting the demand event title and demand event content in the customer service demand data; extracting event element information based on the demand event title and the demand event content, and generating event summary information based on the demand event content; performing event trend analysis on a target event based on the event element information and the event summary information to obtain an event trend analysis result; generating corresponding warning information based on the event trend analysis result, and pushing the warning information to an associated terminal corresponding to a bound associated object. In the customer service complaint data processing solution provided in this application, by comprehensively acquiring and analyzing customer service complaint data and utilizing natural language processing technology, key element information is extracted from the title and content of the complaint event to generate an event summary, thereby improving the accuracy of understanding and analyzing customer issues; trend analysis based on event elements and summary information can monitor and predict potential service problems in real time, predict possible service problems, help companies prepare and respond in advance, and generate early warning information in a timely manner to help companies quickly respond to potential service crises and reduce negative impacts; finally, through automated data processing and early warning information push, efficient and accurate data support is provided to company managers, so that they can make more timely and accurate decisions; in addition, through multi-dimensional data analysis of multiple event systems, a comprehensive perspective is provided to decision makers, thereby achieving comprehensive perception, accurate analysis and real-time early warning of customer service complaints, thereby improving corporate competitiveness. BRIEF DESCRIPTION OF THE DRAWINGS
[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.
[0060] Figure 1 It is an application environment diagram of the customer service demand data processing method provided in an embodiment of the present application;
[0061] Figure 2 It is a flowchart of a method for processing customer service demand data provided in an embodiment of the present application;
[0062] Figure 3 is another flowchart of the customer service demand data processing method provided in an embodiment of the present application;
[0063] Figure 4 is a structural diagram of a customer service demand data processing device provided in an embodiment of the present application;
[0064] Figure 5 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0065] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Instead, they are merely examples of systems and methods consistent with the examples detailed in the attached claims, or with some aspects of the present application.
[0066] It should be noted that, in this article, the terms "include", "comprises" or any other variants thereof are intended to cover descriptions such as non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element. In addition, components, features, and elements with the same name in different embodiments of the present application may have the same meaning or different meanings, and their specific meanings need to be determined by their explanation in the specific embodiment or further combined with the context of the specific embodiment.
[0067] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0068] In the subsequent description, the suffixes such as "module", "component" or "unit" used to represent elements are only used to facilitate the description of the present application, and have no specific meanings. Therefore, "module", "component" or "unit" can be used in a mixed manner.
[0069] The existing intelligent perception and early warning systems for customer service demand situations are mainly based on structured data query statistics and visualization technologies. Data analysis requires a lot of manual intervention, and it is impossible to conduct in-depth and accurate analysis of massive natural language-based customer service data, automatically generate themes or special reports, or make timely warnings for customer service risk events, and it is impossible to provide enterprises with efficient and accurate data support and services.
[0070] In order to solve the above technical problems, the embodiments of the present application provide a customer service demand data processing method, device, electronic device and storage medium. By processing and analyzing a large amount of customer service demand data, comprehensive perception, accurate analysis and real-time warning of customer service demands are achieved, thereby providing more efficient and accurate decision-making support for enterprise management and improving enterprise competitiveness.
[0071] Figure 1 FIG. 1 is an application environment diagram of a method for processing customer service demand data in an embodiment. Figure 1 , the customer service demand data processing method is applied to the customer service demand data processing system. The customer service demand data processing system includes a terminal 110 and a server 120. The terminal 110 and the server 120 are connected through a network. The terminal 110 can be a desktop terminal or a mobile terminal. The mobile terminal can be at least one of a mobile phone, a tablet computer, a laptop computer, etc. The server 120 can be implemented as an independent server or a server cluster composed of multiple servers. The terminal 110 is used to obtain customer service demand data, and the server 120 is used to extract the demand event title and demand event content in the customer service demand data; based on the demand event title and the demand event content, the event element information is extracted, and the event summary information is generated based on the demand event content; based on the event element information and the event summary information, the target event is analyzed for event trend, and the event trend analysis result is obtained; based on the event trend analysis result, the corresponding warning information is generated, and the warning information is pushed to the associated terminal corresponding to the bound associated object.
[0072] See also Figure 2 , Figure 21 is a flow chart of a method for processing customer service demand data provided by an embodiment of the present application. This embodiment mainly uses the application of the method for processing customer service demand data to a computer device as an example to illustrate. The method for processing customer service demand data provided by an embodiment of the present application includes the following steps:
[0073] S1. Obtain customer service demand data, and extract the demand event title and demand event content in the customer service demand data.
[0074] Specifically, for step S1, the main purpose is to obtain customer service demand data from multiple aspects, and then extract the corresponding demand event title and demand event content in the customer service demand data. Among them, the acquisition method can be to obtain customer service demand data from customer service hotlines, emails, social media, online customer service systems, mobile applications, feedback forms, community forums, product evaluations, and customer surveys. In actual embodiments, customers make inquiries or complaints by calling the customer service hotline; customers raise questions or feedback by sending emails; customers may express their demands through social media platforms in public or private messages; customers can communicate with customer service representatives directly on the company's website through online chat tools in real time; some companies provide customer service support through mobile applications, and customers can submit questions or feedback within the application; customers can submit their demands by filling out online forms on the company's website or product; customers may post questions or discussions in community forums operated by the company or third parties; customers leave comments on the product purchase page, which may include feedback on the product or service; companies actively collect customer opinions and suggestions through questionnaires, telephone interviews, etc.
[0075] The acquisition method can be through a web (World Wide Web) interface, which is not specifically limited here, such as using customer service software or CRM systems to automatically record and organize customer interactions and service requests; integrating customer service data from different platforms and channels through application programming interfaces (APIs); using data mining technology to analyze customer feedback from public channels such as social media, forums, and websites; customer service representatives manually recording customer requests and related information after communicating with customers; converting voice calls from telephone customer service to text data for further analysis; performing content analysis on text data such as emails, online chats, and social media posts; evaluating the emotional tendency of customer feedback and identifying positive or negative expressions; aggregating customer feedback from different sources for unified management and analysis; regularly exporting customer service data from various platforms for backup and in-depth analysis; and monitoring the dynamics of customer service channels in real time to quickly respond to customer requests.
[0076] In a specific embodiment, customer service demand information from a customer service hotline, online customer service system, and mobile application can be obtained through a web interface, and information such as event title, content, time, region, source channel, etc. can be extracted and then data fusion can be performed. The fused information can be checked for validity, and errors and duplicate data can be removed to improve data quality.
[0077] S2. Based on the appeal event title and appeal event content, event element information is extracted, and event summary information is generated based on the appeal event content.
[0078] Specifically, for step S2, after extracting the claim event title and claim event content in the customer service claim data, the claim event title and claim event content are combined to extract corresponding event element information, and corresponding event summary information is automatically generated according to the claim event content.
[0079] S3. Perform event trend analysis on the target event based on the event element information and event summary information to obtain event trend analysis results.
[0080] Specifically, for step S3, the target events that meet the analysis conditions are screened out through business rules, so that the target events are subjected to real-time trend analysis and processing based on the extracted event element information and the automatically generated event summary information to predict the future event trends of the target events. The real-time trend analysis and processing can be based on a temporal convolutional network model to perform trend analysis, and finally output the event trend analysis results.
[0081] S4. Based on the event trend analysis results, generate corresponding warning information, and push the warning information to the associated terminal corresponding to the bound associated object.
[0082] Specifically, for step S4, after obtaining the event trend analysis results, the corresponding warning information is automatically generated according to the event trend analysis results, and the warning information is pushed in real time to the associated terminal corresponding to the bound associated object through channels such as SMS, email or WeChat, such as the computer or mobile smart terminal of the enterprise manager, to ensure timely response and disposal.
[0083] It can be seen that the customer service demand data processing method provided in this embodiment, through in-depth and accurate analysis of massive customer service demand data, extracts event element information and automatically generates event summaries, thereby performing trend analysis and early warning on target events based on event element information and event summary information, achieving comprehensive perception, accurate analysis and real-time early warning of customer demands, thereby providing more efficient and accurate decision-making support for enterprise management.
[0084] Optional, such as Figure 3 As shown, in some embodiments, the customer service demand data processing method may further include:
[0085] S5. Generate corresponding analysis reports based on event trend analysis results and warning information;
[0086] S6. Generate a corresponding demand event map based on customer service demand data and event trend analysis results.
[0087] In a specific embodiment, after obtaining the event trend analysis results and warning information, the customer service claim data processing method provided in this embodiment also includes the steps of automatically generating analysis reports, special reports and claim event maps. Based on the event trend analysis results and their corresponding warning information, corresponding daily, weekly and monthly reports can be automatically generated to provide detailed analysis reports. Special reports can also be generated based on specific events to provide support for decision-making. In addition, a corresponding claim event map can be generated in a visual form based on the customer service claim data and event trend analysis results, and the data distribution can be displayed on the claim event map. The event handling results and event trend analysis results can also be displayed, and data interaction is supported. This embodiment uses the event trend analysis results and warning information after model analysis to conduct multi-dimensional visual analysis to provide managers with comprehensive and intuitive analysis results.
[0088] Optionally, in some embodiments, in step S2, “extracting event element information based on the appeal event title and the appeal event content” may specifically include:
[0089] S211. Combine the appeal event title and the appeal event content to obtain a corresponding text combination;
[0090] S212. Use a word segmenter to perform word segmentation processing on the text combination to obtain a first word segmentation result;
[0091] S213. embed the word segmentation results through the pre-trained model as the first model input data;
[0092] S214. Encode the first model input data based on the encoder, and extract corresponding text context-related features based on the encoded first model input data;
[0093] S215. Classify and label each participle through a multi-layer perceptron to obtain the entity type corresponding to each participle;
[0094] S216. Based on the text context-related features and entity types, adjacent word segments of the same entity type are merged to form event element information. The event element information includes the place where the event occurred, the subject involved in the lawsuit, the event content and the event handling result.
[0095] Specifically, the extraction of event element information in step S2 is a key step in the customer service demand data processing method, which ensures that useful information is accurately extracted from the original text to facilitate subsequent analysis and processing. The specific process is as follows:
[0096] Data combination: Combine the title and content of customer service appeal events to form unified text data to ensure the integrity of the context. Word segmentation: Use a word segmenter to segment the combined text and divide the continuous text into independent words or phrases to lay the foundation for subsequent processing. Word embedding: Convert the word segmentation results into word vectors through a pre-trained model. These vectors can capture the semantic information of the words and provide input for the deep learning model. Encoder application: Encode the word vectors using an encoder based on the Transformer architecture (Attention is All You Need model, a deep learning model based on the self-attention mechanism) to extract context-related features in the text. Feature extraction: The encoder can capture the relationship between words and generate feature representations containing rich semantic information. Multilayer Perceptron (MLP): Use MLP to classify and label each word and identify its corresponding entity type, such as location, subject, event content, etc. Entity merging: For adjacent word segments of the same type, merge them to form complete event elements, such as the place where the event occurred and the subject involved in the lawsuit. Formation of core information: Through the above steps, the core information of the incident is formed, including the place where the incident occurred, the parties involved in the lawsuit, the content of the incident and the results of the incident handling.
[0097] Understandably, the advantage of the Transformer architecture lies in its self-attention mechanism, which can better understand long-distance dependencies in text and enhance the model's ability to understand context. In addition, in addition to text data, you can also consider integrating multimodal data such as images and sounds to obtain more comprehensive event information. Over time, the model can continuously update its knowledge base through incremental learning to adapt to new event types and expressions. In the process of extracting key elements, the model's decision-making process should be explainable so that users can understand the basis for the model's judgment. This technology can not only be applied to customer service complaint processing, but can also be extended to other fields, such as legal document analysis, medical record organization, etc. With the development of technology, this process can be processed in real time, providing faster support for event response. Introduce a user feedback mechanism to manually correct the results of model recognition and continuously optimize the accuracy of the model. Expand the model to support multilingual processing capabilities to meet the needs of users who use different languages.
[0098] Optionally, in some embodiments, the “generating event summary information based on the appeal event content” in step S2 may specifically include:
[0099] S221. Use a word segmenter to segment the content of the appeal event to obtain a second word segmentation result;
[0100] S222. embed the word segmentation results through the pre-trained model as the input data of the second model;
[0101] S223. Decoding the second model input data using a decoder, and extracting corresponding text context-related features based on the decoded second model input data;
[0102] S224. Generate a word probability distribution result corresponding to the second model input data through linear transformation and normalized exponential function;
[0103] S225. Calculate the cumulative probability corresponding to each candidate queue in the word probability distribution result by a beam search algorithm;
[0104] S226. Select several candidate queues with the highest cumulative probability as event summary information.
[0105] Specifically, the generation of event summary information in step S2 is the process of converting the long text of customer service appeal event content into concise summaries. These summaries can quickly convey the core information of the event. The specific process is as follows: Word segmentation: Use a word segmenter to decompose the event content text into separate words or phrases to provide a basis for subsequent processing. Word embedding: Convert the word segmentation results into word vectors through a pre-trained model. These vectors can capture the semantic information of the words. Position encoding: Add position encoding to the model input data to maintain the order information of the words in the sequence, which is crucial for capturing contextual relationships. Transformer decoder: Use a decoder based on the Transformer architecture to process the input data. The decoder can understand the contextual features of the text. Feature representation: The decoder extracts feature representations from the text. These features contain the semantic and grammatical information of the text. Linear transformation and Softmax (Softmax Function, a commonly used activation function in machine learning and deep learning): Generate word probability distribution through linear transformation and Softmax function, assign a probability value to each word, indicating the possibility of its appearance in the summary. Beam Search algorithm (a heuristic graph search algorithm): Use the Beam Search algorithm to select from candidate summaries, which evaluates the quality of candidate summaries by cumulative probability. Summary output: Select the k candidate summaries with the highest probability as the final output, which can accurately reflect the main content of the event.
[0106] It is understandable that this embodiment can generate diversified summaries to provide different perspectives on the event by adjusting the parameters of Beam Search. It is possible to process multiple related documents at the same time and generate a comprehensive summary covering multiple texts. An interactive summary system is developed to allow users to specify the length or key information of the summary to customize the summary content. The model is enabled to adapt to texts in different fields and improve the relevance of the summary through field-specific pre-training or fine-tuning. As the event develops, the summary content is updated in real time to reflect the latest progress of the event. An automatic evaluation mechanism is introduced to quantify the quality of the summary and provide feedback to the model for optimization. The model is expanded to support summary generation of multilingual texts to meet the needs of globalization. Allow users to customize the style and focus of the summary as needed, for example, focusing on the impact or cause of the event. Combine text summarization and data visualization to provide a richer way of presenting information. When generating summaries, ensure that personal privacy or information of interest is not disclosed.
[0107] Optionally, in some embodiments, step S3 of "performing an event trend analysis on the target event based on the event element information and the event summary information to obtain an event trend analysis result" may specifically include:
[0108] S31. Determine the target event that meets the analysis conditions through preset business rules;
[0109] S32. Arrange the target events in chronological order, and numerically encode the event element information and event summary information corresponding to the target events to obtain sequence analysis input data;
[0110] S33. Perform feature analysis and event trend analysis on the sequence analysis input data through the temporal convolutional network model to obtain the event trend analysis results corresponding to the target event.
[0111] Specifically, step S3 is an advanced analysis process that comprehensively utilizes data extraction, feature encoding, and deep learning techniques to predict the development trend of events. The specific process is as follows:
[0112] Business rule screening: According to predefined business rules, event information that meets specific analysis conditions is screened from a large amount of event data to ensure the pertinence and accuracy of the analysis. Key element extraction: Extract key elements from the screened events, including time, location, subject, type, etc., to lay the foundation for in-depth analysis. Summary information utilization: Combine the generated event summary to include the main content and key points of the event in the analysis to increase the comprehensiveness of the analysis. Feature element arrangement and encoding: Arrange the extracted feature elements in chronological order, encode these elements numerically, and convert them into a format suitable for machine learning model processing. Sequence analysis input: Form the input data set for sequence analysis, which will be used for training and prediction of the time series prediction model. Time series convolutional network (TCN) application: Use the TCN model to perform feature analysis on sequence data, and use convolutional neural networks to capture local patterns and trends in time series. Dilated convolution: Through the dilated convolution operation, TCN can expand the receptive field, capture longer-term dependencies, and improve the model's prediction ability for time series. Causal convolution layer: Use the causal convolution layer to ensure the causality of the prediction and ensure that the prediction of future events will not be affected by future data. Event frequency and probability prediction: The model outputs predictions on the frequency and probability of future events, providing quantitative risk assessment for corporate decision makers.
[0113] Understandably, more dimensional features can be combined, such as the sentiment and influence of events, to obtain a more comprehensive view of events. Use model fusion technology to combine the prediction results of multiple different models to improve the accuracy and robustness of predictions. Dynamically adjust model parameters according to event types and historical data to adapt to the analysis needs of different event characteristics. Integrate real-time data streams into the analysis process to achieve real-time monitoring and instant prediction of event development trends. Develop interactive visualization tools to enable users to intuitively explore and understand the results of event trend analysis.
[0114] Optionally, in some embodiments, the customer service demand data processing method may further include:
[0115] Perform event clustering based on event element information to obtain repeated appeal events;
[0116] Matching the information of interest based on the content of the appeal event, obtaining a matching result, and determining the event of interest based on the matching result;
[0117] Weak signal analysis is performed based on the content of the appeal event to determine the weak signal event.
[0118] Specifically, before the event trend analysis step is performed on the target event, the target event will also be classified, such as repeated complaint events, events of interest, and weak signal events. For example, through the repeated complaint event detection model, repeated complaint events are detected and annotated; by extracting the information of interest in the event content, the output of whether the time contains the information of interest and the hit keywords of interest is output to determine the event of interest; through the concurrent weak signal analysis model, the weak signals in concurrent events are analyzed to identify potential complaint risks and determine weak signal events.
[0119] Among them, the events of interest in this embodiment can be one or more of customer complaint events, data leakage events, product recall events, service interruption events, negative publicity events, legal problem events, fraud events, supply chain problem events, customer safety problem events, intellectual property infringement events and environmental events. Specifically, customer complaint events refer to formal complaints about product quality, service levels or corporate terms; data leakage events refer to the leakage of customer personal information or transaction data; product recall events refer to large-scale product recalls due to safety issues or other reasons; service interruption events refer to interruptions in services or products that may affect a large number of users; negative publicity events refer to negative information or comments spread through social media or other channels; legal problem events refer to events involving legal proceedings or investigations by third-party agencies; fraud events refer to credit card fraud, service abuse or other types of fraud; supply chain problem events refer to supply chain interruptions or quality issues that affect product delivery; customer safety problem events refer to customer safety issues caused by products or services; intellectual property infringement events refer to infringements involving intellectual property rights; environmental problem events refer to events that have a negative impact on the environment, such as pollution or non-compliant emissions.
[0120] Among them, the concurrent events in this embodiment can be one or more of product defect problem events, service interruption events, price change dispute events, promotional activity problem events, update or change problem events, supply chain problem events, data security events, false advertising events, customer service failure events, legal problem events, technical failure events, and brand reputation crisis events. Specifically, a product defect problem incident refers to a large number of customers reporting defects or functional failures in the same product; a service interruption incident refers to a large-scale interruption of services or products, affecting a wide range of user groups; a price change dispute incident refers to a sudden price change or billing error, which triggers a large number of customer complaints; a promotional activity problem incident refers to unclear promotional activity rules or inconsistent implementation, leading to customer dissatisfaction; an update or change problem incident refers to widespread problems after the software or service is updated, or changes to the terms of service cause customer opposition; a supply chain problem incident refers to supply chain problems causing product delivery delays, affecting a large number of orders; a data security incident refers to data leaks or security vulnerabilities affecting a large number of customers; a false advertising incident refers to advertising that is inconsistent with the actual product or service, triggering concurrent customer complaints; a customer service failure incident refers to customer service's untimely response or improper handling, resulting in dissatisfaction among a large number of customers; a legal problem incident refers to an enterprise's operation violating the law, causing customer concern or litigation; a technical failure incident refers to technical problems causing service unavailability, affecting a large number of users; a brand reputation crisis incident refers to the damage to the brand image due to negative news or events, causing reactions from the customer group.
[0121] Optionally, in some embodiments, step S3 of "performing an event trend analysis on the target event based on the event element information and the event summary information to obtain an event trend analysis result" may specifically include:
[0122] Obtain event element information and event summary information corresponding to repeated appeal events, events of interest, and weak signal events respectively;
[0123] Based on the event element information and event summary information, event trend analysis is performed on repeated demand events, events of interest and weak signal events respectively, and event trend analysis results corresponding to repeated demand events, events of interest and weak signal events are obtained.
[0124] Specifically, the target event may include a repeated demand event, an event of interest, and a weak signal event, and step S3 is specifically as follows: respectively obtaining event element information and event summary information corresponding to the repeated demand event, the event of interest, and the weak signal event; performing event trend analysis on the repeated demand event based on the event element information and event summary information of the repeated demand event, and obtaining an event trend analysis result corresponding to the repeated demand event; performing event trend analysis on the event of interest based on the event element information and event summary information of the event of interest, and obtaining an event trend analysis result corresponding to the event of interest; performing event trend analysis on the weak signal event based on the event element information and event summary information of the weak signal event, and obtaining an event trend analysis result corresponding to the weak signal event.
[0125] Optionally, in some embodiments, the step of “performing event clustering processing based on event element information to obtain repeated appeal events” may specifically include:
[0126] Perform word segmentation and word embedding on the event element information to obtain input data for the third model;
[0127] Input the third model input data into the semantic embedding model, and output the corresponding event vector through the self-attention layer and feedforward neural network of the semantic embedding model;
[0128] Use the cosine similarity function to calculate the similarity between all event vectors and construct a similarity matrix;
[0129] The target events are clustered based on the density clustering algorithm and similarity matrix to obtain repeated appeal events.
[0130] Specifically, for the acquisition method of repeated appeal events, the specific process is as follows: the semantic embedding model SIMCSE based on contrastive learning is used to extract the feature vectors of event element information and cluster the feature vectors. SIMCSE (Semantically Interpretable Matching of Contextualized Embeddings) is a semantic embedding model based on contrastive learning. It learns high-quality semantic representations of texts through pre-training and fine-tuning stages. In the pre-training stage, SIMCSE is first pre-trained on a large amount of unlabeled text, which helps the model learn common language patterns and semantic features. In the fine-tuning stage, fine-tuning is performed using similar text pairs with manual annotations, and the performance of the model on text matching tasks is further improved by comparing data within batches. Through the SIMCSE model, the text related to event element information is converted into feature vectors, which can capture the deep semantic information of the text. The self-attention layer and feedforward neural network in the SIMCSE model work together to generate a comprehensive vector representation for each event. The cosine similarity function is used to calculate the similarity between all event vectors and construct a similarity matrix. In the similarity matrix, according to the preset threshold, the items with similarity higher than the threshold are set to 0, and the other items are set to 1, and converted into a distance matrix. According to the business scenario, such as multiple complaints from the same complainant or repeated events at the same location, the distance matrix is modified to more accurately reflect the similarity between events. The DBSCAN algorithm (Density-Based Spatial Clustering of Applications with Nois) is used to cluster the modified distance matrix to identify and merge the same or similar complaint events. Finally, the DBSCAN algorithm outputs the clustering results, clustering similar complaint events together to provide a basis for further analysis and processing.
[0131] Optionally, in some embodiments, the step of “matching the information of interest based on the content of the appeal event, obtaining a matching result, and determining the event of interest based on the matching result” may specifically include:
[0132] Build a vocabulary of interest based on domain knowledge and experience;
[0133] Based on the word library of interest, a keyword matching method is used to search for keywords in the content of the appeal event as the first information of interest;
[0134] Use a word segmenter to segment the content of the appeal event and remove stop words, and use keyword matching to find keywords in the segmentation results after removing stop words;
[0135] The keywords are feature encoded by the semantic encoding model, and the semantic similarity between the feature-encoded keywords and the vocabulary of interest is calculated by the cosine similarity function;
[0136] Determining second information of interest in the content of the appeal event based on semantic similarity;
[0137] Based on the first information of interest and the second information of interest, it is determined whether the target event is an event of interest, and whether the target event contains the information of interest and the hit keyword of interest.
[0138] Specifically, the extraction of interesting information and the confirmation of interesting events are key links in the analysis of customer service appeal events, which aim to identify and process events that may involve interesting topics or information. The specific process is as follows: First, a vocabulary of interest is constructed. Based on in-depth knowledge and rich experience in a specific field, a vocabulary containing interesting words is constructed. The vocabulary of interest is the basis for identifying information of interest. Using the vocabulary of interest, the information of interest contained in the content of customer service appeal events is searched through keyword matching technology to achieve preliminary screening of information of interest. The event content is segmented and stop words are removed to purify the text data in preparation for keyword extraction. Algorithms such as TF-IDF (TermFrequency-Inverse Document Frequency) and TextRank (graph-based text summarization technology) are applied to extract important keywords in the text, which may indicate information of interest. The semantic encoding model SIMCSE is used to feature encode the extracted keywords and convert the keywords in the text into vector form for semantic comparison. The cosine similarity function is used to calculate the semantic similarity between the text keywords and the words in the vocabulary of interest to determine the degree of proximity between them. According to the similarity threshold, the interesting words in the text that match the interesting word library are identified. These words may indicate the existence of interesting information. The results of keyword matching and semantic matching are integrated to comprehensively judge whether the event content contains interesting information and determine the specific interesting keywords that are hit. Finally, the events containing interesting information are output and the hit interesting keywords are listed to provide a basis for subsequent interesting information processing and decision-making.
[0139] To summarize, the customer service demand data processing method provided in this embodiment comprehensively utilizes natural language processing (NLP) based on deep learning, time series analysis and prediction, data analysis and mining and other technologies to analyze massive customer service complaint data, and timely, comprehensively and accurately discover events that affect enterprise management, enterprise products or enterprise services, thereby solving the pain points of large amounts of various customer service demand data, time-consuming and labor-intensive manual analysis, and untimely analysis results. This embodiment is based on the unified textual structure generation framework UIE (Unified Textual Structure Generation Framework), which extracts core information such as the location of the complaint incident, the subject involved in the complaint, the event content, and the event handling results from the unstructured text; based on the Transformer Decoder (Transformer encoder-decoder) text generation model with the same architecture as chatGPT (ChatGenerative Pre-trained Transformer, a generative pre-trained transformer model for chat), the main content of the complaint is summarized and automatically generated as an event summary; based on the semantic embedding model SIMCSE of contrastive learning, the key information of the complaint is semantically encoded and the similarity is compared; based on the DBSCAN algorithm of density clustering and in combination with expert rules, repeated complaint events are automatically identified and merged; through the BI-GRU (Bidirectional Gated Recurrent Unit, bidirectional gated recurrent unit) architecture, extracts and counts key information and information of interest in the content of the appeal; through multi-dimensional data analysis and TCN time series prediction model, real-time trend analysis is performed on events of interest, multiple events, weak signal events, etc., and early warning information is generated in a timely manner and pushed to enterprise managers through APP or e-mail and other channels for key attention and priority treatment, thereby improving the efficiency and real-time nature of appeal processing.
[0140] In order to facilitate understanding of the customer service demand data processing method provided in this embodiment, this embodiment also provides a customer service demand data processing system for executing the customer service demand data processing method, and the customer service demand data processing system includes a basic support layer, a collection and aggregation layer, a data management layer, a data service layer and a display layer, wherein the basic support layer includes a distributed computing and storage unit, a large-scale parallel processing database, a distributed database and a cloud platform infrastructure; the collection and aggregation layer includes a batch collection unit, an Internet collection unit, an incremental collection unit, a streaming collection unit, a real-time collection unit, a data reporting unit, a data resource directory unit and a data exchange unit, and the data management layer includes a customer service data analysis model and a data resource library , where the customer service data analysis model includes a multidimensional data analysis model, an improper return order model, an event of interest analysis model, a customer service data-specific classification model, a complainant emotion recognition model, an early warning event analysis model, a unified address merging model, an address extraction model, a repeated complaint event detection model, a complaint subject extraction model, and a concurrent weak signal analysis model. The data resource library includes a customer service database, a homogeneous customer service database, and a customer service data analysis result library; the data service layer includes a service catalog unit and a service API unit, and the display layer includes an APP docking unit, a SMS notification unit, an early warning message push unit, a map and block data form interaction unit, a daily / weekly / monthly report generation unit, and a special report generation unit. Taking the customer service demand data processing system to perform data processing and analysis as an example, the specific process is as follows:
[0141] Step 1. Data collection;
[0142] 1.1) Batch collection: Batch obtain customer service complaint data from historical data.
[0143] 1.2) Internet collection: Obtain relevant complaint data in real time through the Internet interface.
[0144] 1.3) Incremental collection: Incrementally update existing data to ensure the real-time and integrity of the data.
[0145] Step 2. Data cleaning, governance, and integration;
[0146] 2.1) Data cleaning: Clean the collected raw data, remove noise and redundant information, and standardize the data format.
[0147] 2.2) Data governance: Apply data governance rules to verify and correct data to ensure data consistency and accuracy.
[0148] 2.3) Data fusion: Integrate data from different channels to form a unified data view.
[0149] Step 3. Data modeling and analysis;
[0150] 3.1) Hotline data analysis: Use a deep learning-based natural language processing (NLP) model to classify and summarize hotline complaint data.
[0151] 3.2) Unified address merging model: merge and standardize address information in different formats.
[0152] 3.3) Address extraction model: extract specific address information from the text.
[0153] 3.4) Repeated complaint event detection model: Detect and annotate repeated complaint events.
[0154] 3.5) Complaint topic extraction model: extract the specific topic and object of the complaint.
[0155] 3.6) Concurrent weak signal analysis model: Analyze weak signals in concurrent events and identify potential risks.
[0156] 3.7) Complainant emotion recognition model: Analyze the complainant’s emotional state through emotion recognition technology.
[0157] Step 4. Multidimensional data analysis;
[0158] 4.1) Multidimensional data analysis model: Perform multidimensional analysis on the cleaned data to explore the implicit relationships and patterns in the data.
[0159] 4.2) Event of Interest Analysis Model: Analyze events of interest to promptly identify potential customer complaint risks.
[0160] 4.3) Warning event analysis model: Analyze and predict potential warning events.
[0161] Step 5. Timing analysis and prediction;
[0162] TCN-based time series prediction model: Use the time series convolutional network (TCN) model to perform real-time trend analysis on events of interest, multiple events, weak signal events, etc., and predict future event trends.
[0163] Step 6. Data display and early warning;
[0164] 6.1) Warning message push: Warning messages are pushed in real time through SMS, WeChat and other channels to ensure timely response and disposal.
[0165] 6.2) Map and data form interaction: Display data distribution on the map and support data form interaction.
[0166] 6.3) Daily / weekly / monthly report generation: Automatically generate daily, weekly and monthly reports and provide detailed analysis reports.
[0167] 6.4) Special report generation: Generate special reports based on specific events to provide data support for enterprises.
[0168] The application scenarios of this embodiment mainly include:
[0169] Repeated complaint analysis: By using AI algorithm models to analyze and merge multiple incidents and counter-complaints, and sorting and displaying them by the number of incidents involved, we can find out the trend of a customer's complaint expanding and fermenting. Timely warnings are sent to the company's responsible departments or managers, providing important reference for the priority of handling customer incidents.
[0170] Enterprise office (AI report writing): Use the large language model to understand and analyze the natural language of a large amount of unstructured customer service data, and automatically write daily reports for different enterprise departments on a regular basis. This solves the problem that grassroots employees spend a lot of manpower on the production of reports of various purposes and formats, and comprehensively and quickly improves the work efficiency of enterprise staff.
[0171] It can be seen that this embodiment uses the self-developed AI (Artificial Intelligence) large language model KuaiGPT (Kuai Generative Pre-trained Transformer) to deeply analyze and mine enterprise risk events and issue early warnings in a timely manner. The model automatically analyzes data and automatically generates special reports. The result data after model analysis is used for multi-dimensional, multi-information cross-cutting, and fine-grained visual analysis to provide comprehensive and intuitive analysis results for enterprise managers.
[0172] It should be understood that although Figure 2 and Figure 3 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2 and Figure 3 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.
[0173] In order to facilitate better implementation of the customer service demand data processing method of the embodiment of the present application, the embodiment of the present invention also provides a customer service demand data processing device based on the above customer service demand data processing method. The meanings of the terms are the same as those in the above customer service demand data processing method, and the specific implementation details can refer to the description in the method embodiment.
[0174] See also Figure 4 , Figure 4 This is a schematic diagram of the structure of a customer service demand data processing device provided in an embodiment of the present application. The customer service demand data processing device includes a data acquisition module 201, a data analysis module 202, an event analysis module 203 and an event warning module 204, wherein:
[0175] The data acquisition module 201 is used to acquire customer service demand data and extract the demand event title and demand event content in the customer service demand data;
[0176] The data analysis module 202 is used to extract event element information based on the appeal event title and the appeal event content, and generate event summary information based on the appeal event content;
[0177] An event analysis module 203 is used to perform event trend analysis on a target event based on event element information and event summary information to obtain an event trend analysis result;
[0178] The event warning module 204 is used to generate corresponding warning information based on the event trend analysis result, and push the warning information to the associated terminal corresponding to the bound associated object.
[0179] For the specific definition of the customer service demand data processing device, please refer to the definition of the customer service demand data processing method above, which will not be repeated here. Each module in the above-mentioned customer service demand data processing device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
[0180] The customer service demand data processing device provided in this embodiment performs in-depth and accurate analysis on massive customer service demand data, extracts event element information and automatically generates event summaries, thereby performing trend analysis and early warning on target events based on the event element information and event summary information, achieving comprehensive perception, accurate analysis and real-time early warning of customer demands, thereby providing more efficient and accurate decision support for enterprise management.
[0181] In addition, the present application also provides an electronic device, such as Figure 5As shown, it shows a schematic diagram of the structure of the electronic device involved in the embodiment of the present application, specifically:
[0182] The electronic device may include components such as a processor 301 with one or more processing cores, a memory 302 with one or more computer-readable storage media, a power supply 303, and an input unit 304. Those skilled in the art will appreciate that Figure 5 The electronic device structure shown in the figure does not constitute a limitation on the electronic device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.
[0183] The processor 301 is the control center of the electronic device. It uses various interfaces and lines to connect various parts of the entire electronic device. By running or executing software programs and / or modules stored in the memory 302, and calling data stored in the memory 302, it executes various functions of the electronic device and processes data, thereby monitoring the electronic device as a whole. Optionally, the processor 301 may include one or more processing cores; preferably, the processor 301 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 301.
[0184] The memory 302 can be used to store software programs and modules. The processor 301 executes various functional applications and customer service demand data processing methods by running the software programs and modules stored in the memory 302. The memory 302 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 302 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory 302 may also include a memory controller to provide the processor 301 with access to the memory 302.
[0185] The electronic device also includes a power supply 303 for supplying power to each component. Preferably, the power supply 303 can be logically connected to the processor 301 through a power management system, so that the power management system can manage charging, discharging, power consumption and other functions. The power supply 303 can also include one or more DC or AC power supplies, recharging systems, power failure detection circuits, power converters or inverters, power status indicators and other arbitrary components.
[0186] The electronic device may further include an input unit 304, which may be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal input related to user settings and function control.
[0187] Although not shown, the electronic device may further include a display unit, etc., which will not be described in detail herein. Specifically in this embodiment, the processor 301 in the electronic device will load the executable files corresponding to the processes of one or more application programs into the memory 302 according to the following instructions, and the processor 301 will run the application programs stored in the memory 302, thereby realizing various functions, as follows:
[0188] Acquire customer service demand data, and extract the demand event title and demand event content in the customer service demand data; extract event element information based on the demand event title and demand event content, and generate event summary information based on the demand event content; perform event trend analysis on the target event based on the event element information and event summary information to obtain event trend analysis results; generate corresponding warning information based on the event trend analysis results, and push the warning information to the associated terminal corresponding to the bound associated object.
[0189] The specific implementation of the above operations can be found in the previous embodiments, which will not be described in detail here.
[0190] The embodiments of the present application perform in-depth and accurate analysis on massive amounts of customer service demand data, extract event element information and automatically generate event summaries, thereby performing trend analysis and early warning on target events based on the event element information and event summary information, achieving comprehensive perception, accurate analysis and real-time early warning of customer demands, thereby providing more efficient and accurate decision support for enterprise management.
[0191] A person of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.
[0192] To this end, an embodiment of the present application provides a storage medium in which a plurality of instructions are stored, and the instructions can be loaded by a processor to execute any step in the customer service demand data processing provided in the embodiment of the present application. For example, the instructions can execute the following steps:
[0193] Acquire customer service demand data, and extract the demand event title and demand event content in the customer service demand data; extract event element information based on the demand event title and demand event content, and generate event summary information based on the demand event content; perform event trend analysis on the target event based on the event element information and event summary information to obtain event trend analysis results; generate corresponding warning information based on the event trend analysis results, and push the warning information to the associated terminal corresponding to the bound associated object.
[0194] The specific implementation of the above operations can be found in the previous embodiments, which will not be described in detail here.
[0195] The storage medium may include: a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.
[0196] Since the instructions stored in the storage medium can execute the steps in any customer service demand data processing method provided in the embodiments of the present application, the beneficial effects that can be achieved by any customer service demand data processing method provided in the embodiments of the present application can be achieved. Please refer to the previous embodiments for details and will not be repeated here.
[0197] The above is a detailed introduction to a customer service demand data processing method, device, electronic device and storage medium provided in an embodiment of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for technical personnel in this field, according to the idea of the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.
Claims
1. A method for processing customer service demand data, characterized in that: The steps include: Acquire customer service demand data, and extract demand event titles and demand event contents from the customer service demand data; Extracting event element information based on the appeal event title and the appeal event content, and generating event summary information based on the appeal event content; Performing event clustering processing based on the event element information to obtain repeated appeal events; Performing weak signal analysis based on the content of the appeal event to determine a weak signal event; Build a vocabulary of interest based on domain knowledge and experience; Based on the interest word library, the keyword matching method is used to search for the keywords in the content of the appeal event as the first interest information; the appeal event content is segmented by a word segmenter and stop words are removed, and the keywords in the segmentation result after the stop words are removed are searched by the keyword matching method; the keywords are feature-encoded by a semantic coding model, and the semantic similarity between the feature-encoded keywords and the interest word library is calculated by a cosine similarity function; the second interest information in the appeal event content is determined based on the semantic similarity; based on the first interest information and the second interest information, it is determined whether the target event is an interest event, and whether the target event contains interest information and hit interest keywords; Performing event trend analysis on the target event based on the event element information and the event summary information to obtain an event trend analysis result; Based on the event trend analysis result, corresponding warning information is generated, and the warning information is pushed to the associated terminal corresponding to the bound associated object.
2. The method for processing customer service demand data according to claim 1, characterized in that: The method further comprises: Based on the event trend analysis results and the warning information, generate a corresponding analysis report; Based on the customer service demand data and the event trend analysis results, a corresponding demand event map is generated.
3. The method for processing customer service demand data according to claim 1, characterized in that: The performing event trend analysis on the target event based on the event element information and the event summary information to obtain an event trend analysis result includes: Respectively obtaining event element information and event summary information corresponding to the repeated appeal event, the event of interest, and the weak signal event; Based on the event element information and the event summary information, event trend analysis is performed on the repeated demand events, the events of interest and the weak signal events respectively to obtain event trend analysis results corresponding to the repeated demand events, the events of interest and the weak signal events.
4. The method for processing customer service demand data according to claim 1, characterized in that: The extracting of event element information based on the appeal event title and the appeal event content includes: Combining the appeal event title and the appeal event content to obtain a corresponding text combination; Using a word segmenter to perform word segmentation processing on the text combination to obtain a first word segmentation result; Perform word embedding on the word segmentation result through a pre-trained model as input data for the first model; Encoding the first model input data based on an encoder, and extracting corresponding text context-related features based on the encoded first model input data; Use a multi-layer perceptron to classify and label each segmentation to obtain the entity type corresponding to each segmentation; Based on the text context-related features and the entity type, adjacent word segments of the same entity type are merged to form event element information, and the event element information includes the place where the event occurred, the subject involved in the lawsuit, the event content and the event handling result.
5. The method for processing customer service demand data according to claim 1, characterized in that: Generating event summary information based on the content of the appeal event, including: Using a word segmenter to perform word segmentation processing on the content of the appeal event to obtain a second word segmentation result; Perform word embedding on the word segmentation result through a pre-trained model as input data for the second model; Using a decoder to decode the second model input data, and extracting corresponding text context-related features based on the decoded second model input data; Generate a word probability distribution result corresponding to the second model input data through linear transformation and normalized exponential function; Calculate the cumulative probability corresponding to each candidate queue in the word probability distribution result by a beam search algorithm; A plurality of candidate queues with the highest cumulative probabilities are selected as event summary information.
6. The method for processing customer service demand data according to claim 1, characterized in that: The performing event trend analysis on the target event based on the event element information and the event summary information to obtain an event trend analysis result includes: Determine target events that meet the analysis criteria through preset business rules; Arranging the target events in chronological order, and numerically encoding the event element information and the event summary information corresponding to the target events to obtain sequence analysis input data; The sequence analysis input data is subjected to feature analysis and event trend analysis through a temporal convolutional network model to obtain an event trend analysis result corresponding to the target event.
7. The method for processing customer service demand data according to claim 1, characterized in that: The event clustering process is performed based on the event element information to obtain repeated appeal events, including: Perform word segmentation and word embedding processing on the event element information to obtain third model input data; Inputting the third model input data into the semantic embedding model, and outputting the corresponding event vector through the self-attention layer and feedforward neural network of the semantic embedding model; Using the cosine similarity function to calculate the similarity between all the event vectors, and construct a similarity matrix; Based on the density clustering algorithm and the similarity matrix, event clustering processing is performed on the target event to obtain repeated appeal events.
8. A customer service demand data processing device, characterized in that: include: A data acquisition module, used to acquire customer service demand data and extract demand event titles and demand event contents from the customer service demand data; A data analysis module, configured to extract event element information based on the appeal event title and the appeal event content, and generate event summary information based on the appeal event content; An event classification module, used for performing event clustering processing based on the event element information to obtain repeated appeal events; Performing weak signal analysis based on the content of the appeal event to determine a weak signal event; Build a vocabulary of interest based on domain knowledge and experience; Based on the interest word library, the keyword matching method is used to search for the keywords in the content of the appeal event as the first interest information; the appeal event content is segmented by a word segmenter and stop words are removed, and the keywords in the segmentation result after the stop words are removed are searched by the keyword matching method; the keywords are feature-encoded by a semantic coding model, and the semantic similarity between the feature-encoded keywords and the interest word library is calculated by a cosine similarity function; the second interest information in the appeal event content is determined based on the semantic similarity; based on the first interest information and the second interest information, it is determined whether the target event is an interest event, and whether the target event contains interest information and hit interest keywords; An event analysis module, used to perform event trend analysis on a target event based on the event element information and the event summary information to obtain an event trend analysis result; The event warning module is used to generate corresponding warning information based on the event trend analysis result, and push the warning information to the associated terminal corresponding to the bound associated object.
9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the customer service demand data processing method as described in any one of claims 1 to 7 are implemented.
10. A storage medium, characterized in that: A computer program is stored which can be loaded by a processor and execute the method for processing customer service demand data as described in any one of claims 1 to 7.
