Optimization method and system of intelligent customer service system, equipment and medium

Abstract

The invention is suitable for the field of intelligent customer service, and discloses an optimization method, system and device of an intelligent customer service system and a medium. Performing natural language processing on the dialogue content; according to the personalized characteristics of the user, adjusting the word segmentation spacing and the boundary of the semantic unit, and generating an optimized word segmentation and semantic analysis result; an emotion analysis model is applied to the optimized word segmentation and semantic analysis result, the emotion state of the user is recognized, and an emotion recognition result is obtained; identifying the question type of the user by adopting a classification model; and generating a reply suggestion in combination with the emotional state and the question type of the user. By identifying the personalized characteristics of the user and adjusting the word segmentation spacing and the semantic unit boundary of the dialogue content, not only can the coherence of expressing brief user semantic understanding be improved, but also the accuracy of redundant expression user semantic analysis can be enhanced, so that a more accurate word segmentation and semantic analysis result is generated.

Description

Technical Field

[0001] The present invention relates to the field of intelligent customer service, and in particular to an optimization method, system, equipment and medium for an intelligent customer service system. Background Art

[0002] In current intelligent customer service systems, natural language processing (NLP) technology has become the key to improving user experience and service quality. However, existing intelligent customer service systems often use a unified word segmentation and semantic analysis method when processing user conversations, ignoring the individual differences between different users, such as language habits and expression styles. This "one-size-fits-all" approach may perform well when processing short or standardized user inputs, but when faced with complex, personalized or lengthy user expressions, it is often difficult to accurately capture the user's true intentions, resulting in a decrease in the accuracy of word segmentation and semantic analysis, which in turn affects subsequent sentiment analysis and problem identification, and ultimately generates less accurate or thoughtful reply suggestions. Summary of the invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes an optimization method, system, device and medium for an intelligent customer service system.

[0004] In a first aspect, the present application proposes an optimization method for an intelligent customer service system, comprising: obtaining conversation content input by a user through a front end of a customer service system;

[0005] Performing natural language processing on the acquired conversation content, wherein the natural language processing includes word segmentation, part-of-speech tagging, and semantic analysis;

[0006] According to the personalized characteristics of users, adjust the word segmentation spacing and the boundaries of semantic units to generate optimized word segmentation and semantic analysis results;

[0007] Applying a sentiment analysis model to the optimized word segmentation and semantic analysis results to identify the user's emotional state and obtain sentiment recognition results;

[0008] Use classification models to identify the user's question type;

[0009] A response suggestion is generated based on the user's emotional state and the question type.

[0010] More specifically, in the above technical solution, the word segmentation spacing and the boundaries of the semantic units are adjusted according to the personalized characteristics of the user, specifically including:

[0011] Adjust the word segmentation spacing according to the user's personalized characteristics, including expression habits;

[0012] If the user's expression is brief, shorten the word segmentation interval;

[0013] If the user's expression is lengthy, increase the word spacing;

[0014] Perform part-of-speech tagging on the adjusted word segmentation results to determine the part of speech of each word;

[0015] A semantic analysis algorithm is used to perform semantic analysis on the part-of-speech tagging results, identify semantic units, and obtain semantic understanding results.

[0016] More specifically, in the above technical solution, adjusting the word segmentation spacing and the boundaries of the semantic units according to the personalized characteristics of the user also includes:

[0017] Based on the results of word segmentation and part-of-speech tagging, semantic units are preliminarily identified, and the semantic units include phrases, clauses or sentence fragments;

[0018] Analyze the context information of the current conversation content to obtain the user's current intention and focus, wherein the context information includes the previous and next sentences and the topic;

[0019] By understanding the semantics of words, we can determine which words are classified into the same semantic unit. The semantics of words include synonyms, antonyms, and hyponyms.

[0020] More specifically, in the above technical solution, before generating the reply suggestion, it also includes:

[0021] Obtaining historical user interaction data, evaluating the user's knowledge level by analyzing the vocabulary complexity and frequency of professional terms used in the user's questions in combination with a knowledge base, wherein the knowledge base is used to retrieve relevant reply suggestions;

[0022] According to the knowledge level of the user, the keyword matching strategy during knowledge base search is adjusted. If the user has a high knowledge level, the weight of professional terms is increased; if the user has a low knowledge level, the weight of popular expressions is increased;

[0023] Using the adjusted keyword matching strategy, searching the knowledge base for text segments related to the user's question;

[0024] Perform natural language processing on the search results, use the text summarization model to extract information, and generate preliminary answer content.

[0025] More specifically, in the above technical solution, after the preliminary answer content is generated, the preliminary answer content is input into the sentiment analysis model to identify the sentiment tendency of the preliminary answer content;

[0026] The preliminary answer content is adjusted according to the user's knowledge level. If the knowledge level is high, more professional expressions are retained; if the knowledge level is low, popular explanations are added;

[0027] At the same time, the tone and emotional tendency of the answer content are adjusted according to the sentiment analysis results.

[0028] More specifically, in the above technical solution, when generating reply suggestions, the recommendation system based on the collaborative filtering algorithm combines the user's emotional state, question type, user portrait, question solving history and knowledge base search results to generate reply suggestions; specifically including:

[0029] Based on the user's registration information, browsing history, purchase history and search history, obtain the user's personal information, interest preferences and historical behavior, and build a user profile;

[0030] Problem solving history is the user's past problems and their solutions, including successful and failed solutions, to analyze the user's problem solving path and preferences;

[0031] Building a knowledge base, through which relevant suggestions are retrieved when a user asks a question;

[0032] Obtain the user's emotional state code, question type code, user portrait feature vector and question solving history features, and use collaborative filtering algorithm to calculate the similarity between users or questions;

[0033] Find other users who are similar to the target user based on user similarity, and analyze the response suggestions of the other users in similar situations in the past;

[0034] Based on the similarity of the questions, find other questions similar to the target question, and generate multiple reply suggestions according to the solutions corresponding to the other questions and the current user's question;

[0035] A comprehensive evaluation is performed on multiple reply suggestions, the reply suggestions are sorted according to the evaluation results, and the best reply suggestion is selected and sent to the user.

[0036] More specifically, the above technical solution also includes: obtaining user feedback on the reply suggestion, and adjusting the optimization method of the intelligent customer service system based on the feedback.

[0037] In a second aspect, the present application also proposes an optimization system for an intelligent customer service system, comprising:

[0038] The first processing module: obtains the conversation content input by the user through the front end of the customer service system;

[0039] The second processing module performs natural language processing on the acquired conversation content, wherein the natural language processing includes word segmentation, part-of-speech tagging and semantic analysis;

[0040] The third processing module: adjusts the word segmentation spacing and the boundaries of semantic units according to the user's personalized characteristics, and generates optimized word segmentation and semantic analysis results;

[0041] Fourth processing module: applying a sentiment analysis model to the optimized word segmentation and semantic analysis results to identify the user's emotional state and obtain emotion recognition results;

[0042] The fifth processing module: using the classification model to identify the user's question type;

[0043] The sixth processing module: generates a reply suggestion based on the user's emotional state and the question type.

[0044] In a third aspect, the present application further proposes a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any of the methods described above when executing the computer program.

[0045] In a fourth aspect, the present application further proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method as described in any one of the above items is implemented.

[0046] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0047] When the user's expression is relatively lengthy and contains multiple information points, the system will separate related but semantically independent words or phrases by widening the word segmentation distance and semantic unit boundaries, avoiding semantic confusion caused by excessive vocabulary aggregation. For example, when a user inputs "What is the weather forecast for tomorrow, including information such as temperature, humidity, and wind speed?", the system can accurately divide words such as "temperature", "humidity", and "wind speed" into different semantic units, ensuring accurate understanding of each information point. The system will also use semantic analysis algorithms to deeply understand the semantic relationships between words (such as synonyms, antonyms, hyponyms, etc.), so as to more accurately determine which words should be classified into the same semantic unit and which should be distinguished, further improving the accuracy of semantic understanding.

[0048] For users with a higher level of knowledge, the system increases the weight of professional terms, so that when searching the knowledge base, professional answers related to the user's questions can be matched first. This strategy meets the needs of high-level users for professionalism and accuracy. For users with a lower level of knowledge, the system increases the weight of popular expressions, making the search results easier to understand. This strategy reduces the difficulty for users to understand the reply content and improves the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 It is a flowchart of an optimization method of an intelligent customer service system provided by an embodiment of the present invention;

[0051] Figure 2 It is a structural diagram of an optimization system of an intelligent customer service system provided by an embodiment of the present invention;

[0052] Figure 3 It is a schematic diagram of the structure of the terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0053] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0054] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0055] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0056] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0057] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0058] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0059] See also Figure 1 The present application proposes an optimization method for an intelligent customer service system, comprising: S101 obtaining the conversation content input by the user through the front end of the customer service system; S102 performing natural language processing on the obtained conversation content, wherein the natural language processing includes word segmentation, part-of-speech tagging and semantic analysis; S103 adjusting the word segmentation spacing and the boundary of the semantic unit according to the personalized characteristics of the user, and generating optimized word segmentation and semantic analysis results; S104 applying a sentiment analysis model to the optimized word segmentation and semantic analysis results to identify the user's emotional state and obtain sentiment recognition results; S105 using a classification model to identify the user's question type; S106 generating a reply suggestion in combination with the user's emotional state and the question type.

[0060] When the user's expression is relatively brief and the intention is clear, the system reduces the distance between word segments and divides closely related words into the same semantic unit, thereby reducing semantic breaks caused by improper word segmentation and improving the coherence of semantic understanding. For example, when a user enters "What will the weather be like tomorrow?", the system can accurately regard "tomorrow's weather" as a whole semantic unit, avoiding the mistaken splitting of it into "tomorrow" and "weather", which would lead to a deviation in understanding. The system also combines the contextual information of the conversation, such as the previous and next sentences, topics, etc., to further understand the user's current intentions and concerns, and ensure semantic coherence throughout the conversation.

[0061] Specifically, first, the conversation content between the user and the customer service system is preprocessed by natural language processing technology. The conversation content is segmented using the ji eba word segmentation tool, and the stop words are removed using the Harbin Institute of Technology stop word list, and then the THULAC tool is used for part-of-speech tagging. Then, semantic analysis technologies such as Word2Vec are used to convert the preprocessed conversation content into a 300-dimensional semantic vector. Next, the LSTM-based sentiment analysis model is used to perform emotion recognition on the semantic vector, and the accuracy of this model in the emotion classification task can reach 85%. According to the semantic vector and emotional state, the SVM classification model is used to identify the user's question type, and the classification accuracy reaches more than 90%. When the user is identified as having a negative emotion, the system retrieves relevant reply suggestions from the knowledge base, and uses the collaborative filtering algorithm to dynamically adjust the retrieval weight according to the user's portrait characteristics, such as age, gender, purchase history, etc., to improve the personalization of the recommendation. If the user is in a positive emotion, the standard reply is retrieved, and the retrieval parameters are also dynamically adjusted according to the user's characteristics.

[0062] Optionally, the search results are returned to the user, and user feedback is collected. The sentiment analysis and question classification models are optimized in real time using reinforcement learning algorithms to continuously improve system performance. The optimized models are applied to subsequent user conversations to form a closed-loop feedback mechanism, which continuously improves the system intelligence level and user experience satisfaction.

[0063] In some embodiments, when the customer service system performs word segmentation and semantic analysis on the user's conversation content, it dynamically adjusts the spacing and boundaries between word segmentation and semantic units according to the user's personalized characteristics. For example, for users who express themselves concisely and clearly, the system can appropriately reduce the spacing between word segmentation and semantic units to improve the coherence of semantic understanding; for users who express themselves verbosely and lengthily, the system can appropriately increase the spacing between word segmentation and semantic units to improve the accuracy of semantic understanding.

[0064] According to the personalized characteristics of the user, such as expression habits, the word segmentation spacing is dynamically adjusted. If the user's expression is brief, the word segmentation spacing is shortened; if the user's expression is lengthy, the word segmentation spacing is widened; the adjusted word segmentation results are tagged with parts of speech to determine the part of speech of each word in preparation for semantic analysis; a semantic analysis algorithm is used to perform semantic analysis on the part-of-speech tagging results, identify semantic units, and obtain semantic understanding results; based on the results of word segmentation and part-of-speech tagging, the system preliminarily identifies possible semantic units, which include phrases, clauses, or longer sentence fragments; the system needs to analyze the context information of the current conversation, including the preceding and following sentences, topics, etc., to understand the user's current intentions and concerns; by understanding the semantic relationships between words (such as synonyms, antonyms, hyponyms, etc.), the system can more accurately determine which words should be classified into the same semantic unit.

[0065] In some embodiments, before generating reply suggestions, the parameters and weights of the knowledge base retrieval algorithm are dynamically adjusted according to the user's knowledge level (obtained by analyzing the user's historical interaction data and the complexity of the question) to generate targeted retrieval results. For users with rich domain knowledge, the system can appropriately increase the matching degree of professional terms; for users with insufficient domain knowledge, the system can appropriately reduce the matching degree of professional terms and increase the matching degree of popular expressions.

[0066] Obtain user historical interaction data, evaluate the user's knowledge level by analyzing the vocabulary complexity and frequency of professional terms used in the user's questions; dynamically adjust the keyword matching strategy during knowledge base retrieval according to the user's knowledge level, if the user's knowledge level is high, increase the weight of professional terms, if the user's knowledge level is low, increase the weight of popular expressions; use the adjusted keyword matching strategy to retrieve text fragments related to the user's questions in the knowledge base; perform natural language processing on the retrieval results, use a text summary model to extract key information, and generate preliminary answer content; input the preliminary answer content into a sentiment analysis model to identify the sentiment tendency contained in the answer; adjust the preliminary answer content according to the user's knowledge level, if the knowledge level is high, retain more professional expressions, if the knowledge level is low, add popular explanations; at the same time, adjust the tone and sentiment tendency of the answer according to the sentiment analysis results; return the optimized answer content to the user, record the interaction data of this time, and update the user knowledge level evaluation model; dynamically adjust the knowledge base retrieval strategy and answer generation strategy according to the user's feedback and questions in this interaction.

[0067] Specifically, in order to obtain historical user interaction data, the system will record the content, timestamp and other information of each question asked by the user and store them in the MongoDB database. Through natural language processing of the user's question content, the TF-IDF algorithm is used to calculate the weight of the keywords in the question, and the complexity of the question is comprehensively evaluated by combining indicators such as question length and vocabulary richness. According to the complexity score, users are divided into three knowledge levels: elementary, intermediate and advanced. The system will dynamically adjust the parameters of the Elasticsearch search engine. If the user's knowledge level is advanced, the weight of professional terms will be increased by 20%, while the weight of popular expressions will be reduced; otherwise, the weight of professional terms will be reduced by 20% and the weight of popular expressions will be increased by 10%. The adjusted retrieval algorithm will be matched in the knowledge base, which contains 5 million structured domain knowledge data. The search results are semantically analyzed, the BERT model is used to extract key information, and the Seq2Seq model is used to generate preliminary answers. The system will also identify the user's emotional state through the sentiment analysis model, and the emotional state is divided into three categories: positive, neutral and negative. Finally, the preliminary answer is optimized based on the user's emotions and knowledge level. If the user is positive and has a high level of knowledge, the proportion of professional terms in the answer is appropriately increased. If the user is negative and has a low level of knowledge, soothing language is added to the answer. The optimized answer is returned to the user through the API interface. At the same time, the interaction data will be recorded to update the user knowledge level assessment model and optimize the retrieval strategy for the next interaction. Through this series of processes, the system can provide personalized domain knowledge Q&A services based on the user's actual situation and improve the user experience.

[0068] In some embodiments, the recommendation system based on the collaborative filtering algorithm combines the user's emotional state, question type, user profile information, question solving history and knowledge base search results to generate personalized reply suggestions. Specifically, it includes:

[0069] Through natural language processing (NLP) technology, such as sentiment analysis, emotion labels (such as positive, negative, neutral, etc.) are extracted from the user's historical conversations or text input, and the sentiment dictionary or pre-trained sentiment analysis model is used to analyze the user's input emotions in real time; Named Entity Recognition (NER) and intent recognition in NLP technology are used to parse user questions and classify them into specific types (such as query, suggestion, complaint, etc.); collect user personal information, interest preferences, historical behavior, etc., and build a comprehensive user portrait. The user portrait data comes from user registration information, browsing history, purchase history, search history, etc.; record users' past problems and their solutions, including successful and failed cases, and analyze users' problem-solving paths and preferences so that historical experience can be considered when making recommendations; establish a knowledge base containing common questions and their answers. When users ask questions, they first search through the knowledge base to find relevant answers or suggestions. Extract key features from the above data, such as user emotional state encoding, question type encoding, user portrait feature vector, problem solving history features, etc. For text data, models such as TF-IDF, Word2Vec or BERT can be used for vectorized representation. Based on the extracted features, a collaborative filtering algorithm is used to calculate the similarity between users or questions. Common similarity calculation methods include cosine similarity, Pearson correlation coefficient, Jaccard similarity coefficient, etc.; find other users similar to the target user, and analyze the reply suggestions of these users in similar situations in the past; treat user questions as "items", calculate the similarity between different questions, find questions similar to the target question, and generate reply suggestions based on the solutions corresponding to these questions and the specific situation of the current user; combine knowledge base retrieval results, user portrait information, emotional state and other factors to comprehensively evaluate multiple candidate reply suggestions; sort the reply suggestions according to the evaluation results, and select the best reply suggestion to present to the user.

[0070] It should be noted that the user's emotional state is a reflection of the user's current psychological state, which can be identified from the user's text input or historical conversations through sentiment analysis technology; after extracting the emotional state, it is usually necessary to encode it into a computer-understandable format, such as using digital labels (such as 1 for positive, 0 for negative, and -1 for neutral) or more complex sentiment vectors (including scores of multiple emotional dimensions). The question type refers to the category or intention of the question raised by the user. Natural language processing technology (such as named entity recognition, intent recognition, etc.) can be used to parse user questions and classify them into predefined question types (such as query, suggestion, complaint, etc.); similarly, the question type also needs to be encoded into a computer-understandable format, such as using digital labels or one-hot encoding. User portrait is an abstract representation of comprehensive user information, including multiple dimensions such as user personal information, interest preferences, and historical behavior. When extracting user portrait features, this information needs to be converted into numerical features, and feature engineering (such as feature selection, feature scaling, feature combination, etc.) may be used to optimize the feature representation; ultimately, user portrait features will be organized into a feature vector for use in subsequent recommendation algorithms. Problem solving history records users' past problems and their solutions, including successful and failed cases. When extracting problem solving history features, we can focus on user problem type, solution time, solution method (such as self-service solution, customer service assistance, etc.), solution result (such as success, failure), and other dimensions; these features can be used to analyze users' problem solving preferences and habits, thereby providing more personalized suggestions when making recommendations.

[0071] In some embodiments, obtaining user feedback on the reply suggestion and adjusting the optimization method of the intelligent customer service system based on the feedback specifically includes:

[0072] Get user feedback on reply suggestions and adjust the sentiment analysis model and recommendation system accordingly. Set a feedback button on the customer service system interface. After clicking it, a form pops up and users can fill in satisfaction ratings, text comments, etc. Provide star ratings or emoji ratings below the reply suggestions to allow users to quickly express their satisfaction. In the instant chat interface, users can directly enter feedback in the chat box.

[0073] The user feedback data is collected through the designed feedback mechanism. The data includes key information such as user satisfaction score, comment content, timestamp, user ID, etc. At the same time, contextual information such as conversation content related to the feedback, reply suggestions, etc. needs to be recorded for subsequent analysis.

[0074] Conduct in-depth analysis of the collected feedback data to identify problems and improvement points in the system. The analysis may include the following aspects:

[0075] Sentiment analysis: Use sentiment analysis techniques to identify the emotional tendencies in user reviews, such as positive, negative, or neutral.

[0076] Topic extraction: Use natural language processing technology to extract topics or keywords from user comments to understand the main issues that users are concerned about.

[0077] Trend analysis: Analyze feedback data over time to identify fluctuations in system performance and potential problems.

[0078] According to the analysis results of the feedback data, formulate specific optimization strategies. The optimization strategy can be carried out in the following aspects:

[0079] Improve reply suggestions: Adjust the generation logic of reply suggestions based on user feedback to improve the accuracy and personalization of replies.

[0080] Optimize natural language processing: Improve the accuracy and efficiency of natural language processing technologies such as word segmentation, part-of-speech tagging, and semantic analysis.

[0081] Improve user experience: Optimize system interface and interaction process to improve user convenience and satisfaction.

[0082] Implement the formulated optimization strategy and make corresponding adjustments and improvements to the intelligent customer service system, including modifying the code, updating the model, adjusting parameters, etc.

[0083] See also Figure 2 , the present application also proposes an optimization system for an intelligent customer service system, comprising:

[0084] The first processing module 201: obtaining the conversation content input by the user through the front end of the customer service system;

[0085] Second processing module 202: performing natural language processing on the acquired conversation content, wherein the natural language processing includes word segmentation, part-of-speech tagging and semantic analysis;

[0086] The third processing module 203: adjusting the word segmentation spacing and the boundaries of the semantic units according to the personalized characteristics of the user, and generating optimized word segmentation and semantic analysis results;

[0087] Fourth processing module 204: applying a sentiment analysis model to the optimized word segmentation and semantic analysis results to identify the user's emotional state and obtain an emotion recognition result;

[0088] The fifth processing module 205: using the classification model to identify the type of user's question;

[0089] The sixth processing module 206: generates a reply suggestion based on the user's emotional state and the question type.

[0090] It is understandable that if Figure 1 The contents of the optimization method embodiment of the intelligent customer service system shown in the figure are applicable to the optimization system embodiment of the intelligent customer service system. The functions specifically implemented by the optimization system embodiment of the intelligent customer service system are similar to those in the example shown in the figure. Figure 1 The optimization method embodiment of the intelligent customer service system shown in FIG. 1 is the same as that of FIG. 1 , and the beneficial effects achieved are the same as those of FIG. Figure 1 The beneficial effects achieved by the optimization method embodiment of the intelligent customer service system shown are also the same.

[0091] It should be noted that the information interaction, execution process and other contents between the above-mentioned systems are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0092] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0093] See also Figure 3 The embodiment of the present invention further provides a computer device 3, comprising: a memory 302 and a processor 301 and a computer program 303 stored in the memory 302. When the computer program 303 is executed on the processor 301, an optimization method for an intelligent customer service system as described in any one of the above methods is implemented.

[0094] The computer device 3 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The computer device 3 may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art will appreciate that Figure 3 It is only an example of computer device 3 and does not constitute a limitation on computer device 3. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components, for example, it may also include input and output devices, network access devices, etc.

[0095] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0096] In some embodiments, the memory 302 may be an internal storage unit of the computer device 3, such as a hard disk or memory of the computer device 3. In other embodiments, the memory 302 may also be an external storage device of the computer device 3, such as a plug-in hard disk, a smart memory card (Smart Med iaCard, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 3. Further, the memory 302 may also include both an internal storage unit of the computer device 3 and an external storage device. The memory 302 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program, etc. The memory 302 may also be used to temporarily store data that has been output or is to be output.

[0097] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the optimization method of the intelligent customer service system as described in any one of the above methods is implemented.

[0098] In this embodiment, if the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can at least include: any entity or device that can carry the computer program code to the camera / terminal device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0099] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. An optimization method for an intelligent customer service system, characterized in that: include: Obtain the conversation content entered by the user through the front end of the customer service system; Performing natural language processing on the acquired conversation content, wherein the natural language processing includes word segmentation, part-of-speech tagging, and semantic analysis; According to the personalized characteristics of users, adjust the word segmentation spacing and the boundaries of semantic units to generate optimized word segmentation and semantic analysis results; Applying a sentiment analysis model to the optimized word segmentation and semantic analysis results to identify the user's emotional state and obtain sentiment recognition results; Use classification models to identify the type of user questions; A response suggestion is generated based on the user's emotional state and the question type.

2. The optimization method of the intelligent customer service system according to claim 1, characterized in that: According to the personalized characteristics of users, adjust the word spacing and the boundaries of semantic units, including: Adjust the word segmentation spacing according to the user's personalized characteristics, including expression habits; If the user's expression is brief, shorten the word segmentation interval; If the user's expression is lengthy, increase the word spacing; Perform part-of-speech tagging on the adjusted word segmentation results to determine the part of speech of each word; A semantic analysis algorithm is used to perform semantic analysis on the part-of-speech tagging results, identify semantic units, and obtain semantic understanding results.

3. The optimization method of the intelligent customer service system according to claim 2, characterized in that: According to the user's personalized characteristics, adjust the word spacing and the boundaries of semantic units, including: Based on the results of word segmentation and part-of-speech tagging, semantic units are preliminarily identified, and the semantic units include phrases, clauses or sentence fragments; Analyze the context information of the current conversation content to obtain the user's current intention and focus, wherein the context information includes the previous and next sentences and the topic; By understanding the semantics of words, we can determine which words are classified into the same semantic unit. The semantics of words include synonyms, antonyms, and hyponyms.

4. The optimization method of the intelligent customer service system according to claim 1, characterized in that: Before generating reply suggestions, also include: Obtaining historical user interaction data, evaluating the user's knowledge level by analyzing the vocabulary complexity and frequency of professional terms used in the user's questions in combination with a knowledge base, wherein the knowledge base is used to retrieve relevant reply suggestions; According to the knowledge level of the user, the keyword matching strategy during knowledge base search is adjusted. If the user has a high knowledge level, the weight of professional terms is increased; if the user has a low knowledge level, the weight of popular expressions is increased; Using the adjusted keyword matching strategy, searching the knowledge base for text fragments related to the user's question; Perform natural language processing on the search results, use the text summarization model to extract information, and generate preliminary answer content.

5. The optimization method of the intelligent customer service system according to claim 4, characterized in that: After generating the preliminary answer content, inputting the preliminary answer content into the sentiment analysis model to identify the sentiment tendency of the preliminary answer content; The preliminary answer content is adjusted according to the user's knowledge level. If the knowledge level is high, more professional expressions are retained; if the knowledge level is low, popular explanations are added; At the same time, the tone and emotional tendency of the answer content are adjusted according to the sentiment analysis results.

6. The optimization method of the intelligent customer service system according to claim 1, characterized in that: When generating reply suggestions, the recommendation system based on the collaborative filtering algorithm combines the user's emotional state, question type, user profile, problem solving history and knowledge base search results to generate reply suggestions; Specifically include: Based on the user's registration information, browsing history, purchase history and search history, obtain the user's personal information, interest preferences and historical behavior, and build a user profile; Problem solving history is the user's past problems and their solutions, including successful and failed solutions, to analyze the user's problem solving path and preferences; Building a knowledge base, through which relevant suggestions are retrieved when a user asks a question; Obtain the user's emotional state code, question type code, user portrait feature vector and question solving history features, and use collaborative filtering algorithm to calculate the similarity between users or questions; Find other users who are similar to the target user based on user similarity, and analyze the response suggestions of the other users in similar situations in the past; Based on the similarity of the questions, find other questions similar to the target question, and generate multiple reply suggestions according to the solutions corresponding to the other questions and the current user's question; A comprehensive evaluation is performed on multiple reply suggestions, the reply suggestions are sorted according to the evaluation results, and the best reply suggestion is selected and sent to the user.

7. The optimization method of the intelligent customer service system according to claim 1, characterized in that: Also includes: Obtain user feedback on the reply suggestion, and adjust the optimization method of the intelligent customer service system based on the feedback.

8. An optimization system for an intelligent customer service system, characterized in that: include: The first processing module: obtains the conversation content input by the user through the front end of the customer service system; The second processing module performs natural language processing on the acquired conversation content, wherein the natural language processing includes word segmentation, part-of-speech tagging and semantic analysis; The third processing module: adjusts the word segmentation spacing and the boundaries of semantic units according to the user's personalized characteristics, and generates optimized word segmentation and semantic analysis results; Fourth processing module: applying a sentiment analysis model to the optimized word segmentation and semantic analysis results to identify the user's emotional state and obtain emotion recognition results; The fifth processing module: using the classification model to identify the user's question type; The sixth processing module: generates a reply suggestion based on the user's emotional state and the question type.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images