Intelligent customer service method and system based on large model

Abstract

The embodiment of the invention provides an intelligent customer service method and system based on a large model. The method comprises the following steps: firstly, receiving input information of a user in real time through an input module to ensure that the user can initiate interaction in a most convenient way; then, the sentiment analysis module performs deep analysis on the input information by means of a pre-trained large model, identifies the sentiment state of the user and generates a corresponding sentiment label, thereby laying a foundation for subsequent humanized reply; then, a context understanding module extracts context information including dialogue context logic and user intentions based on historical interaction records of the user, the problem that a traditional system is insufficient in understanding of complex contexts is solved, and the system can better meet the user requirements; and finally, the reply generation module selects an optimal reply strategy from a preset reply strategy library in combination with the emotion label and the context information, and generates target reply content through a large model, so that the pertinence of reply is ensured, and the reply can accord with the dialogue context and the user intention, thereby remarkably improving the user experience.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to an intelligent customer service method and system based on a large model. Background Technology

[0002] With the rapid development of internet and mobile communication technologies, the way businesses and customers interact has undergone tremendous changes. Intelligent customer service systems, as a crucial tool for communication between businesses and customers, have become an indispensable part of many industries.

[0003] Currently, traditional intelligent customer service systems mainly rely on keyword matching and preset rules to process user requests.

[0004] However, existing intelligent customer service systems rely on keyword matching and preset rules to process user requests, and often lack effective identification and response mechanisms for user emotions, resulting in a poor user experience. Summary of the Invention

[0005] The embodiments of the present invention provide an intelligent customer service method and system based on a large model, which can improve the user experience.

[0006] In a first aspect, embodiments of the present invention provide an intelligent customer service method based on a large model, the method comprising:

[0007] The input module receives the current user's input information in real time, wherein the input information includes text and voice.

[0008] Based on a pre-trained large model, sentiment analysis is performed on the current user's input information to identify the current user's emotional state and generate corresponding sentiment labels;

[0009] Based on the current user's historical interaction records, the historical interaction records are modeled using a context understanding module, and context information is extracted, wherein the context information includes: dialogue context logic and user intent;

[0010] Based on the sentiment tags and the context information, the response generation module extracts the target response information from a pre-established response strategy library and outputs it to the current user.

[0011] Preferably,

[0012] After receiving the current user's input information in real time using the input module, and before performing sentiment analysis on the current user's input information based on the pre-trained large model to identify the current user's emotional state and generate corresponding sentiment labels, the method further includes:

[0013] The preprocessing module performs data preprocessing operations on the current user's input information to generate target input information. The data preprocessing operations include data cleaning and formatting. Data cleaning includes removing noise, punctuation marks, and stop words from the current user's input information. Formatting converts the current user's input information into a standard data format.

[0014] The pre-trained large model performs sentiment analysis on the current user's input information to identify the current user's emotional state and generates corresponding sentiment labels, including:

[0015] Based on a pre-trained large model, sentiment analysis is performed on the target input information to identify the current user's emotional state and generate corresponding sentiment tags.

[0016] Preferably,

[0017] The pre-trained large model performs sentiment analysis on the current user's input information to identify the current user's emotional state and generates corresponding sentiment labels, including:

[0018] Based on a pre-trained large model, text features of the target input information are extracted using word embedding and sentence embedding techniques and input into the sentiment detection model. The text features include semantic information, sentiment expression level, frequency, and context sentiment tendency association.

[0019] The sentiment analysis of the text features is performed using a sentiment detection model to identify the current user's emotional state.

[0020] After identifying the current user's emotional state, a corresponding emotional tag is generated based on the current user's emotional state;

[0021] Based on the degree and frequency of emotional expression and the correlation with the contextual emotional tendency in the text features, the emotional intensity corresponding to each identified emotional state is evaluated, and an emotional intensity evaluation result including the emotional intensity level is generated.

[0022] The step of extracting target reply information from a pre-established reply strategy library using the reply generation module based on the sentiment tag and the context information, and outputting it to the current user, includes:

[0023] Based on the sentiment tag, the sentiment intensity assessment result, and the context information, the response generation module extracts the target response information from the pre-established response strategy library and outputs it to the current user.

[0024] Preferably,

[0025] After extracting target reply information from a pre-established reply strategy library using the reply generation module based on the sentiment tag and the context information, and outputting it to the current user, the method further includes:

[0026] The user feedback module collects feedback information from the current user regarding the target response information, wherein the feedback information includes: satisfaction rating and improvement suggestions;

[0027] The feedback information is categorized and processed, and analyzed using a sentiment analysis algorithm to generate the current user's sentiment feedback;

[0028] Based on the current user's emotional feedback and the historical interaction records, a training dataset is generated, wherein the historical interaction records include: user input information, dialogue history, and response interaction records. The training dataset is used to optimize the parameters of a large model.

[0029] Based on the training dataset, the parameters of the large model are updated using an incremental learning algorithm, wherein the large model includes: the sentiment detection model, the context understanding model, and the response generation model.

[0030] Secondly, embodiments of the present invention provide an intelligent customer service system based on a large model, the system comprising:

[0031] Input module: The user receives input information from the current user in real time, wherein the input information includes text and voice;

[0032] Sentiment analysis module: Based on a pre-trained large model, the module performs sentiment analysis on the input information of the current user to identify the current user's emotional state and generate corresponding sentiment tags;

[0033] Context understanding module: used to model the historical interaction records of the current user and extract context information, wherein the context information includes: dialogue context logic and user intent;

[0034] Response generation module: Based on the sentiment tag and the context information, it extracts target response information from a pre-established response strategy library and outputs it to the current user.

[0035] Preferably,

[0036] Following the input module and preceding the sentiment analysis module, the system further includes a preprocessing module;

[0037] The preprocessing module performs data preprocessing operations on the current user's input information to generate target input information. The data preprocessing operations include data cleaning and formatting. Data cleaning includes removing noise, punctuation marks, and stop words from the current user's input information. Formatting converts the current user's input information into a standard data format.

[0038] The sentiment analysis module is also used to perform sentiment analysis on the target input information based on a pre-trained large model to identify the current user's emotional state and generate corresponding sentiment tags.

[0039] Preferably,

[0040] The sentiment analysis module is also used to perform:

[0041] Based on a pre-trained large model, text features of the target input information are extracted using word embedding and sentence embedding techniques and input into the sentiment detection model. The text features include semantic information, sentiment expression level, frequency, and context sentiment tendency association.

[0042] The sentiment analysis of the text features is performed using a sentiment detection model to identify the current user's emotional state.

[0043] After identifying the current user's emotional state, a corresponding emotional tag is generated based on the current user's emotional state;

[0044] Based on the degree and frequency of emotional expression and the correlation with the contextual emotional tendency in the text features, the emotional intensity corresponding to each identified emotional state is evaluated, and an emotional intensity evaluation result including the emotional intensity level is generated.

[0045] Based on the emotional state, corresponding emotional tags are generated, and the emotional intensity corresponding to each emotional state is evaluated based on the emotional detection model.

[0046] The response generation module is further configured to extract target response information from a pre-established response strategy library and output it to the current user based on the emotion tag, the emotion intensity assessment result, and the context information.

[0047] Preferably,

[0048] Following the response generation module, the system further includes: a user feedback module;

[0049] The user feedback module is used to perform:

[0050] Collect feedback information from the current user regarding the target response information, wherein the feedback information includes: satisfaction rating and improvement suggestions;

[0051] The feedback information is categorized and processed, and analyzed using a sentiment analysis algorithm to generate the current user's sentiment feedback;

[0052] Based on the current user's emotional feedback and the historical interaction records, a training dataset is generated, wherein the historical interaction records include: user input information, dialogue history, and response interaction records. The training dataset is used to optimize the parameters of a large model.

[0053] Based on the training dataset, the parameters of the large model are updated using an incremental learning algorithm, wherein the large model includes: the sentiment detection model, the context understanding model, and the response generation model.

[0054] Thirdly, embodiments of the present invention provide an intelligent customer service system based on a large model, comprising: at least one memory and at least one processor;

[0055] The at least one memory is used to store a machine-readable program;

[0056] The at least one processor is configured to invoke the machine-readable program to execute any of the methods described in the first aspect.

[0057] Fourthly, embodiments of the present invention provide a computer-readable medium storing computer instructions, which, when executed by a processor, cause the processor to perform any of the methods described in the first aspect.

[0058] This invention provides an intelligent customer service method and system based on a large model. The method first receives text or voice input from users in real time through an input module, ensuring users can initiate interactions in the most convenient way. Then, a sentiment analysis module uses a pre-trained large model to perform in-depth analysis of the input information, accurately identifying the user's emotional state and generating corresponding sentiment tags, laying the foundation for subsequent personalized responses. Next, a contextual understanding module relies on the user's historical interaction records to extract contextual information, including dialogue logic and user intent, solving the problem of traditional systems' insufficient understanding of complex contexts and allowing the system to better meet user needs. Finally, a response generation module combines sentiment tags and contextual information to select the optimal response strategy from a pre-set response strategy library, and then generates the target response content through the large model. This ensures the response is targeted and matches the dialogue context and user intent, thereby significantly improving the user experience. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a flowchart of an intelligent customer service method based on a large model provided in an embodiment of the present invention;

[0061] Figure 2 This is a flowchart of another intelligent customer service method based on a large model provided in an embodiment of the present invention;

[0062] Figure 3 This is a schematic diagram of an intelligent customer service system based on a large model provided in an embodiment of the present invention;

[0063] Figure 4 This is a schematic diagram of another intelligent customer service system based on a large model provided in an embodiment of the present invention;

[0064] Figure 5 This is a schematic diagram of another intelligent customer service system based on a large model provided in an embodiment of the present invention. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0066] like Figure 1 As shown, this embodiment of the invention provides an intelligent customer service method based on a large model, which may include the following steps:

[0067] Step 101: Receive the current user's input information in real time using the input module. The input information includes text and voice.

[0068] Step 102: Based on the pre-trained large model, perform sentiment analysis on the current user's input information to identify the current user's emotional state and generate corresponding sentiment labels;

[0069] Step 103: Based on the current user's historical interaction records, use the context understanding module to model the historical interaction records and extract context information, which includes: dialogue context logic and user intent;

[0070] Step 104: Based on sentiment tags and contextual information, use the response generation module to extract target response information from a pre-established response strategy library and output it to the current user.

[0071] This invention provides an intelligent customer service method based on a large model. The method first receives text or voice input from users in real time through an input module, ensuring users can initiate interactions in the most convenient way. Then, a sentiment analysis module uses a pre-trained large model to perform in-depth analysis of the input information, accurately identifying the user's emotional state and generating corresponding sentiment tags, laying the foundation for subsequent personalized responses. Next, a contextual understanding module relies on the user's historical interaction records to extract contextual information, including dialogue logic and user intent, solving the problem of insufficient understanding of complex contexts in traditional systems and allowing the system to better meet user needs. Finally, a response generation module combines sentiment tags and contextual information to select the optimal response strategy from a pre-set response strategy library, and then generates the target response content through the large model. This ensures the response is targeted and matches the dialogue context and user intent, thereby significantly improving the user experience.

[0072] To provide high-quality data for subsequent analysis, in one embodiment of the present invention, after step 101 and before step 102, the following is further included:

[0073] The preprocessing module performs data preprocessing operations on the current user's input information to generate target input information. The data preprocessing operations include data cleaning and formatting. Data cleaning includes removing noise, punctuation marks, and stop words from the current user's input information. Formatting converts the current user's input information into a standard data format.

[0074] Step 102 in the above embodiments may specifically include the following steps:

[0075] Based on a pre-trained large model, sentiment analysis is performed on the target input information to identify the current user's emotional state and generate corresponding sentiment tags.

[0076] In this embodiment of the invention, in order to provide high-quality data for subsequent analysis, the preprocessing module first performs data cleaning (e.g., noise removal, punctuation removal, stop words removal, etc.) and formatting (e.g., converting to a uniform format) on the text or speech converted from user input to generate target input information, ensuring the consistency and accuracy of the target input information. On this basis, the sentiment analysis module uses a pre-trained large model to perform in-depth analysis on the target input information, and uses a sentiment detection model to identify the user's emotional state (e.g., anger, dissatisfaction, happiness, etc.) in real time and generate corresponding sentiment tags. This process makes full use of the advantages of the large model in sentiment perception, allowing the system to accurately capture the user's emotional fluctuations, providing key basis for generating responses that fit the user's emotional needs, and thus improving the overall experience through service interactions that better understand user emotions.

[0077] To optimize the overall service experience, in one embodiment of the present invention, step 102 in the above embodiment may specifically include the following steps:

[0078] Based on a pre-trained large model, text features of the target input information are extracted using word embedding and sentence embedding techniques and input into the sentiment detection model. The text features include semantic information, sentiment expression level, frequency, and context sentiment tendency association.

[0079] The sentiment analysis of the text features is performed using a sentiment detection model to identify the current user's emotional state.

[0080] After identifying the current user's emotional state, a corresponding emotional tag is generated based on the current user's emotional state;

[0081] Based on the degree and frequency of emotional expression and the correlation with the contextual emotional tendency in the text features, the emotional intensity corresponding to each identified emotional state is evaluated, and an emotional intensity evaluation result including the emotional intensity level is generated.

[0082] Step 104 in the above embodiments may specifically include the following steps:

[0083] Based on the sentiment tag, the sentiment intensity assessment result, and the context information, the response generation module extracts the target response information from the pre-established response strategy library and outputs it to the current user.

[0084] In this embodiment of the invention, firstly, text features are extracted from the preprocessed target input information using word embedding and sentence embedding techniques from a pre-trained large model, transforming the text into a computer-understandable vector representation and capturing the semantic and emotional connotations behind the language. Next, a deep learning framework is used to train an emotion detection model, employing a cross-entropy loss function and the Adam optimizer for model training. This allows for in-depth analysis of these text features, accurately identifying the user's current emotional state (e.g., anger, dissatisfaction, happiness). Subsequently, corresponding emotion tags are generated based on the identification results, and the model evaluates the emotional intensity corresponding to each user's emotional state. This process ensures the accuracy of emotion recognition by relying on the powerful semantic understanding capabilities of the large model, and achieves a refined grasp of the user's emotions through emotion intensity evaluation. This provides a more detailed basis for the subsequent response generation module, enabling the system to adjust its response strategy according to the type and intensity of the user's emotions, making the response more aligned with the user's current psychological state. This significantly improves the user's sense of being understood and satisfaction during the interaction process, optimizing the overall service experience.

[0085] To continuously optimize system performance, in one embodiment of the present invention, after step 104, the above embodiment further includes:

[0086] The user feedback module collects feedback information from the current user regarding the target response information, wherein the feedback information includes: satisfaction rating and improvement suggestions;

[0087] The feedback information is categorized and processed, and analyzed using a sentiment analysis algorithm to generate the current user's sentiment feedback;

[0088] Based on the current user's emotional feedback and the historical interaction records, a training dataset is generated, wherein the historical interaction records include: user input information, dialogue history, and response interaction records. The training dataset is used to optimize the parameters of a large model.

[0089] Based on the training dataset, the parameters of the large model are updated using an incremental learning algorithm, wherein the large model includes: the sentiment detection model, the context understanding model, and the response generation model.

[0090] In this embodiment of the invention, model parameters can be updated periodically based on user emotional feedback and historical interaction records to continuously optimize system performance. Through the above system architecture, large-scale models and sentiment analysis technology can be effectively integrated to achieve accurate perception of user emotions and personalized services. The user feedback and model optimization process is a key mechanism for the system to achieve adaptive learning, continuously improve service quality, and optimize user experience. A feedback loop mechanism is established to promptly feed user feedback into the model training and system optimization processes. The user feedback module collects user feedback information, such as satisfaction ratings and improvement suggestions for the target response (e.g., a user-friendly feedback collection interface can be designed). It then uses a text classification model to categorize the feedback and employs sentiment analysis algorithms to mine user emotional feedback, accurately grasping the user's overall experience with the service. Subsequently, a training dataset is built based on users' historical interaction records, providing rich and real-world data support for model optimization. Finally, an incremental learning algorithm is used to gradually update the model parameters of the sentiment detection model, context understanding model, and response generation model based on this training dataset. This avoids retraining the entire model, allowing the system to continuously adapt to changing user needs, gradually improving sentiment recognition accuracy, context understanding depth, and response content relevance, ultimately providing users with a more expected and personalized service experience. Furthermore, A / B testing can be conducted regularly to compare the performance of different system versions and select the best solution for deployment.

[0091] In one embodiment of the invention, by integrating advanced large-scale models and sentiment analysis algorithms, the system can accurately perceive the user's emotional state in real time and adjust its response strategy accordingly, providing a more humanized service experience, effectively alleviating the user's emotional problems, and improving user satisfaction. Utilizing the powerful contextual understanding and multi-turn dialogue management capabilities of the large-scale model, the system can accurately understand the user's true needs in complex dialogue scenarios, providing more precise and effective solutions, and significantly improving the system's performance in handling multi-task and multi-turn dialogues. Based on deep learning of user historical interaction records and behavioral patterns, the system can provide customized service suggestions for each user, better meeting their specific needs and enhancing user loyalty and satisfaction with the brand. Simultaneously, the system possesses continuous learning and self-optimization capabilities, extracting valuable information from each user interaction to continuously improve service strategies, enabling the system to quickly adapt and provide high-quality service when facing new problems and special situations. Through the comprehensive application of sentiment perception, contextual understanding, personalized service, and adaptive learning, the system can provide a more intelligent, flexible, and humanized customer service experience, thereby improving user satisfaction and corporate image. Furthermore, efficient sentiment perception and contextual understanding capabilities also reduce the need for manual intervention, improve the automation level of customer service, and reduce the company's operating costs. The system can respond to user needs quickly and accurately, shorten problem-solving time, improve service efficiency, and enable enterprises to gain an advantage in a highly competitive market.

[0092] like Figure 2 As shown, to more clearly illustrate the technical solution and advantages of the present invention, the following provides a detailed description of the intelligent customer service method based on a large model, which may include the following steps:

[0093] Step 201: Receive the current user's input information in real time using the input module, whereby the input information includes text and voice.

[0094] Specifically, text input from users can be received through an API interface, and the user's speech can be converted into text using speech recognition technology.

[0095] Step 202: Use the preprocessing module to perform data preprocessing operations on the current user's input information to generate target input information. The data preprocessing operations include data cleaning and formatting. Data cleaning includes removing noise, punctuation marks, and stop words from the current user's input information. Formatting converts the current user's input information into a standard data format.

[0096] Step 203: Based on the pre-trained large model, extract text features of the target input information using word embedding and sentence embedding techniques and input them into the sentiment detection model. The text features include semantic information, sentiment expression level, frequency, and context sentiment tendency association.

[0097] Step 204: Use a sentiment detection model to perform sentiment analysis on text features and identify the current user's emotional state;

[0098] Step 205: After identifying the current user's emotional state, generate corresponding emotional tags based on the current user's emotional state;

[0099] Step 206: Based on the correlation between the degree and frequency of emotional expression in the text features and the sentiment tendency of the context, evaluate the emotional intensity corresponding to each identified emotional state and generate an emotional intensity evaluation result including the emotional intensity level;

[0100] Step 207: Based on the current user's historical interaction records, use the context understanding module to model the historical interaction records and extract context information, which includes: dialogue context logic and user intent;

[0101] Specifically, efficient data structures (such as graph databases and relational databases) can be designed to store users' dialogue history. Sequence-to-sequence models can be used to model the dialogue history and extract contextual information. At the same time, intent recognition models can be combined to identify users' intents and needs.

[0102] Step 208: Based on sentiment tags, sentiment intensity assessment results, and contextual information, use the response generation module to extract target response information from a pre-established response strategy library and output it to the current user;

[0103] Specifically, a variety of response strategy libraries can be established, covering response templates for different emotional states and contextual scenarios. Rule-based and machine learning-based strategy selection algorithms can be designed to select the optimal response strategy based on sentiment tags and contextual information. A large model can then be used to generate specific response content. The generated response content is further optimized through post-processing steps such as grammatical correction and sentiment enhancement.

[0104] Step 209: Use the user feedback module to collect feedback information from current users regarding the target response information, including: satisfaction rating and improvement suggestions;

[0105] Step 210: Classify the feedback information and analyze it using a sentiment analysis algorithm to generate the current user's sentiment feedback;

[0106] Step 211: Generate a training dataset based on the current user's emotional feedback and historical interaction records. The historical interaction records include: user input information, dialogue history, and response interaction records. The training dataset is used to optimize the parameters of the large model.

[0107] Step 212: Based on the training dataset, update the parameters of the large model using an incremental learning algorithm. The large model includes: sentiment detection model, context understanding model, and response generation model.

[0108] like Figure 3 As shown, this embodiment of the invention provides an intelligent customer service system based on a large model, the system comprising:

[0109] Input module 301: The user receives input information from the current user in real time, wherein the input information includes: text and voice;

[0110] Sentiment analysis module 302: Based on a pre-trained large model, performs sentiment analysis on the input information of the current user to identify the current user's emotional state and generate corresponding sentiment tags;

[0111] Context understanding module 303: is used to model the historical interaction records based on the current user's historical interaction records and extract context information, wherein the context information includes: dialogue context logic and user intent;

[0112] Response generation module 304: Based on the sentiment tag and the context information, it extracts target response information from a pre-established response strategy library and outputs it to the current user.

[0113] based on Figure 3 The intelligent customer service system based on a large model shown is, for example... Figure 4 As shown, after the input module 301 and before the sentiment analysis module 302, a preprocessing module 305 is further included;

[0114] The preprocessing module 305 performs data preprocessing operations on the current user's input information to generate target input information. The data preprocessing operations include data cleaning and formatting. The data cleaning includes removing noise, punctuation marks, and stop words from the current user's input information. The formatting converts the current user's input information into a standard data format.

[0115] The sentiment analysis module 302 is also used to perform sentiment analysis on the target input information based on a pre-trained large model to identify the current user's emotional state and generate corresponding sentiment tags.

[0116] like Figure 4 As shown, the sentiment analysis module 302 is also used to perform:

[0117] Based on a pre-trained large model, text features of the target input information are extracted using word embedding and sentence embedding techniques and input into the sentiment detection model. The text features include semantic information, sentiment expression level, frequency, and context sentiment tendency association.

[0118] The sentiment analysis of the text features is performed using a sentiment detection model to identify the current user's emotional state.

[0119] After identifying the current user's emotional state, a corresponding emotional tag is generated based on the current user's emotional state;

[0120] Based on the degree and frequency of emotional expression and the correlation with the contextual emotional tendency in the text features, the emotional intensity corresponding to each identified emotional state is evaluated, and an emotional intensity evaluation result including the emotional intensity level is generated.

[0121] Based on the emotional state, corresponding emotional tags are generated, and the emotional intensity corresponding to each emotional state is evaluated based on the emotional detection model.

[0122] The response generation module 304 is further configured to extract target response information from a pre-established response strategy library and output it to the current user based on the emotion tag, the emotion intensity assessment result, and the context information.

[0123] like Figure 4 As shown, after the response generation module 304, a user feedback module 306 is further included;

[0124] The user feedback module 306 is used to perform:

[0125] Collect feedback information from the current user regarding the target response information, wherein the feedback information includes: satisfaction rating and improvement suggestions;

[0126] The feedback information is categorized and processed, and analyzed using a sentiment analysis algorithm to generate the current user's sentiment feedback;

[0127] Based on the current user's emotional feedback and the historical interaction records, a training dataset is generated, wherein the historical interaction records include: user input information, dialogue history, and response interaction records. The training dataset is used to optimize the parameters of a large model.

[0128] Based on the training dataset, the parameters of the large model are updated using an incremental learning algorithm, wherein the large model includes: the sentiment detection model, the context understanding model, and the response generation model.

[0129] It is understood that the structures illustrated in the embodiments of the present invention do not constitute a specific limitation on the intelligent customer service system based on a large model. In other embodiments of the present invention, the intelligent customer service system based on a large model may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0130] The information interaction and execution process between the various units in the above-mentioned device are based on the same concept as the method embodiment of the present invention, and the specific details can be found in the description of the method embodiment of the present invention, and will not be repeated here.

[0131] This invention also provides an intelligent customer service system based on a large model, comprising: at least one memory and at least one processor;

[0132] At least one memory for storing machine-readable programs;

[0133] At least one processor is configured to invoke a machine-readable program to execute the intelligent customer service method based on a large model according to any embodiment of the present invention.

[0134] This invention also provides a computer-readable medium storing computer instructions, which, when executed by a processor, cause the processor to perform the intelligent customer service method based on a large model according to any embodiment of this invention.

[0135] Specifically, a system or apparatus equipped with a storage medium may be provided, on which software program code implementing the functions of any of the embodiments described above is stored, and the computer (or CPU or MPU) of the system or apparatus may read and execute the program code stored in the storage medium.

[0136] In this case, the program code read from the storage medium can itself implement the function of any of the above embodiments, and therefore the program code and the storage medium storing the program code constitute part of the present invention.

[0137] Examples of storage media used to provide program code include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, program code can be downloaded from a server computer via a communication network.

[0138] Furthermore, it should be clear that not only can the program code read by the computer be executed, but also the operating system or other components operating on the computer can be instructed based on the program code to perform some or all of the actual operations, thereby realizing the function of any of the embodiments described above.

[0139] Furthermore, it is understood that the program code read from the storage medium is written to the memory set in the expansion board inserted into the computer or to the memory set in the expansion unit connected to the computer. Then, based on the instructions of the program code, the CPU or other components installed on the expansion board or expansion unit execute some and all of the actual operations, thereby realizing the function of any of the embodiments described above.

[0140] The various embodiments of the present invention have at least the following beneficial effects:

[0141] 1. In this embodiment of the invention, an intelligent customer service method based on a large model is provided. The method first receives text or voice input from users in real time through an input module, ensuring users can initiate interactions in the most convenient way. Then, a sentiment analysis module uses a pre-trained large model to perform in-depth analysis of the input information, accurately identifying the user's emotional state and generating corresponding sentiment tags, laying the foundation for subsequent personalized responses. Next, a contextual understanding module relies on the user's historical interaction records to extract contextual information, including dialogue context logic and user intent, solving the problem of insufficient understanding of complex contexts in traditional systems and allowing the system to better meet user needs. Finally, a response generation module combines sentiment tags and contextual information to select the optimal response strategy from a preset response strategy library, and then generates the target response content through the large model, ensuring both the relevance of the response and its alignment with the dialogue context and user intent, thereby significantly improving the user experience.

[0142] 2. In this embodiment of the invention, in order to provide high-quality data for subsequent analysis, the preprocessing module needs to perform data cleaning and formatting on the text or speech converted from user input to generate target input information, ensuring the consistency and accuracy of the target input information. On this basis, the sentiment analysis module uses a pre-trained large model to perform in-depth analysis on the target input information, and uses the sentiment detection model to identify the user's emotional state in real time and generate corresponding sentiment tags. This process makes full use of the advantages of the large model in sentiment perception, allowing the system to accurately capture the user's emotional fluctuations, providing key basis for generating responses that fit the user's emotional needs, and thus improving the overall experience through service interactions that better understand user emotions.

[0143] 3. In this embodiment of the invention, firstly, text features are extracted from the preprocessed target input information using word embedding and sentence embedding techniques of a pre-trained large model, transforming the text into a computer-understandable vector representation and capturing the semantic and emotional connotations behind the language. Next, a deep learning framework is used to train an emotion detection model, employing a cross-entropy loss function and the Adam optimizer for model training. This allows for in-depth analysis of these text features, accurately identifying the user's current emotional state. Subsequently, corresponding emotion tags are generated based on the identification results, and the model evaluates the emotional intensity corresponding to each user's emotional state. This process not only relies on the powerful semantic understanding capabilities of the large model to ensure the accuracy of emotion recognition but also achieves a refined grasp of the user's emotions through emotion intensity evaluation. This provides a more detailed basis for the subsequent response generation module, enabling the system to adjust its response strategy according to the type and intensity of the user's emotions, making the response more aligned with the user's current psychological state. This significantly improves the user's sense of being understood and satisfaction during the interaction process, optimizing the overall service experience.

[0144] It should be noted that not all steps and modules in the above processes and system structure diagrams are mandatory; some steps or modules can be omitted as needed. The execution order of each step is not fixed and can be adjusted as required. The system structure described in the above embodiments can be a physical structure or a logical structure. That is, some modules may be implemented by the same physical entity, or some modules may be implemented by multiple physical entities, or they may be jointly implemented by certain components in multiple independent devices.

[0145] In the above embodiments, the hardware units can be implemented mechanically or electrically. For example, a hardware unit may include permanent dedicated circuitry or logic (such as a dedicated processor, FPGA, or ASIC) to perform the corresponding operation. The hardware unit may also include programmable logic or circuitry (such as a general-purpose processor or other programmable processor), which can be temporarily configured by software to perform the corresponding operation. The specific implementation method (mechanical, dedicated permanent circuitry, or temporarily configured circuitry) can be determined based on cost and time considerations.

[0146] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A large-scale model-based intelligent customer service method, characterized in that: The method includes: The input module receives the current user's input information in real time, wherein the input information includes text and voice. Based on a pre-trained large model, sentiment analysis is performed on the current user's input information to identify the current user's emotional state and generate corresponding sentiment labels; Based on the current user's historical interaction records, the historical interaction records are modeled using a context understanding module, and context information is extracted, wherein the context information includes: dialogue context logic and user intent; Based on the sentiment tags and the context information, the response generation module extracts the target response information from a pre-established response strategy library and outputs it to the current user.

2. The method according to claim 1, characterized in that, After receiving the current user's input information in real time using the input module, and before performing sentiment analysis on the current user's input information based on the pre-trained large model to identify the current user's emotional state and generate corresponding sentiment labels, the method further includes: The preprocessing module performs data preprocessing operations on the current user's input information to generate target input information. The data preprocessing operations include data cleaning and formatting. Data cleaning includes removing noise, punctuation marks, and stop words from the current user's input information. Formatting converts the current user's input information into a standard data format. The pre-trained large model performs sentiment analysis on the current user's input information to identify the current user's emotional state and generates corresponding sentiment labels, including: Based on a pre-trained large model, sentiment analysis is performed on the target input information to identify the current user's emotional state and generate corresponding sentiment tags.

3. The method according to claim 2, characterized in that, The pre-trained large model performs sentiment analysis on the current user's input information to identify the current user's emotional state and generates corresponding sentiment labels, including: Based on a pre-trained large model, text features of the target input information are extracted using word embedding and sentence embedding techniques and input into the sentiment detection model. The text features include semantic information, sentiment expression level, frequency, and context sentiment tendency association. The sentiment analysis of the text features is performed using a sentiment detection model to identify the current user's emotional state. After identifying the current user's emotional state, a corresponding emotional tag is generated based on the current user's emotional state; Based on the degree and frequency of emotional expression and the correlation with the contextual emotional tendency in the text features, the emotional intensity corresponding to each identified emotional state is evaluated, and an emotional intensity evaluation result including the emotional intensity level is generated. The step of extracting target reply information from a pre-established reply strategy library using the reply generation module based on the sentiment tag and the context information, and outputting it to the current user, includes: Based on the sentiment tag, the sentiment intensity assessment result, and the context information, the response generation module extracts the target response information from the pre-established response strategy library and outputs it to the current user.

4. The method according to any one of claims 1-3, characterized in that, After extracting target reply information from a pre-established reply strategy library using the reply generation module based on the sentiment tag and the context information, and outputting it to the current user, the method further includes: The user feedback module collects feedback information from the current user regarding the target response information, wherein the feedback information includes: satisfaction rating and improvement suggestions; The feedback information is categorized and processed, and analyzed using a sentiment analysis algorithm to generate the current user's sentiment feedback; Based on the current user's emotional feedback and the historical interaction records, a training dataset is generated, wherein the historical interaction records include: user input information, dialogue history, and response interaction records. The training dataset is used to optimize the parameters of a large model. Based on the training dataset, the parameters of the large model are updated using an incremental learning algorithm, wherein the large model includes: the sentiment detection model, the context understanding model, and the response generation model.

5. An intelligent customer service system based on a large model, characterized in that: The system includes: Input module: The user receives input information from the current user in real time, wherein the input information includes text and voice; Sentiment analysis module: Based on a pre-trained large model, the module performs sentiment analysis on the input information of the current user to identify the current user's emotional state and generate corresponding sentiment tags; Context understanding module: used to model the historical interaction records of the current user and extract context information, wherein the context information includes: dialogue context logic and user intent; Response generation module: Based on the sentiment tag and the context information, it extracts target response information from a pre-established response strategy library and outputs it to the current user.

6. The system according to claim 5, characterized in that, Following the input module and preceding the sentiment analysis module, the system further includes a preprocessing module; The preprocessing module performs data preprocessing operations on the current user's input information to generate target input information. The data preprocessing operations include data cleaning and formatting. Data cleaning includes removing noise, punctuation marks, and stop words from the current user's input information. Formatting converts the current user's input information into a standard data format. The sentiment analysis module is also used to perform sentiment analysis on the target input information based on a pre-trained large model to identify the current user's emotional state and generate corresponding sentiment tags.

7. The system according to claim 6, characterized in that, The sentiment analysis module is also used to perform: Based on a pre-trained large model, text features of the target input information are extracted using word embedding and sentence embedding techniques and input into the sentiment detection model. The text features include semantic information, sentiment expression level, frequency, and context sentiment tendency association. The sentiment analysis of the text features is performed using a sentiment detection model to identify the current user's emotional state. After identifying the current user's emotional state, a corresponding emotional tag is generated based on the current user's emotional state; Based on the degree and frequency of emotional expression and the correlation with the contextual emotional tendency in the text features, the emotional intensity corresponding to each identified emotional state is evaluated, and an emotional intensity evaluation result including the emotional intensity level is generated. Based on the emotional state, corresponding emotional tags are generated, and the emotional intensity corresponding to each emotional state is evaluated based on the emotional detection model. The response generation module is further configured to extract target response information from a pre-established response strategy library and output it to the current user based on the emotion tag, the emotion intensity assessment result, and the context information.

8. The system according to any one of claims 5-7, characterized in that, Following the response generation module, the system further includes: a user feedback module; The user feedback module is used to perform: Collect feedback information from the current user regarding the target response information, wherein the feedback information includes: satisfaction rating and improvement suggestions; The feedback information is categorized and processed, and analyzed using a sentiment analysis algorithm to generate the current user's sentiment feedback; Based on the current user's emotional feedback and the historical interaction records, a training dataset is generated, wherein the historical interaction records include: user input information, dialogue history, and response interaction records. The training dataset is used to optimize the parameters of a large model. Based on the training dataset, the parameters of the large model are updated using an incremental learning algorithm, wherein the large model includes: the sentiment detection model, the context understanding model, and the response generation model.

9. An intelligent customer service system based on a large model, characterized in that: include: At least one memory and at least one processor; The at least one memory is used to store a machine-readable program; The at least one processor is configured to invoke the machine-readable program to perform the method according to any one of claims 1 to 4.

10. A computer-readable medium, characterized in that, The computer-readable medium stores computer instructions that, when executed by a processor, cause the processor to perform the method described in any one of claims 1 to 4.

Images