Intelligent interactive customer service system based on AI large model

Abstract

The invention specifically relates to the technical field of natural language processing, and discloses an intelligent interactive customer service system based on an AI large model, which comprises a multi-modal input analysis module, an intention preliminary detection module, a large model semantic understanding module, a dynamic knowledge base association module, a dialogue strategy generation module, a natural language generation module and a service feedback work order module, through deep cooperation of a multi-modal input analysis module and a large model semantic understanding module, entity and attribute information is accurately extracted, context semantic association information is mined, and through the cooperative effect of an intention preliminary detection module, a dynamic knowledge base association module and a dialogue strategy generation module, based on a semantic analysis result output by a large model, a dialogue strategy is generated. According to the method and the system, matching calculation is carried out with the constructed dynamic knowledge base, rich knowledge support is provided for solving complex problems, personalized knowledge content is recommended for different users through the natural language generation module and the service feedback work order module, and the reply matching degree and the complex problem processing capacity are greatly improved.

Description

Technical Field

[0001] This invention relates to the field of natural language processing technology, and more specifically, to an intelligent interactive customer service system based on a large AI model. Background Technology

[0002] With the advancement of business digitalization and the increasing demands for consumer service experience, customer service, as a core link, urgently needs to transform from traditional manual service to intelligent and automated service. In order to improve customer service efficiency and reduce operating costs, intelligent customer service centers have become a key solution. With the development of natural language processing technology, it has provided strong technical support for the development of intelligent interactive systems.

[0003] Existing intelligent interactive customer service systems include a rules engine module, a knowledge base module, a voice recognition module, a dialogue management module, and a work order system module. They have the advantages of instant response, stable service, data analysis to assist decision-making, and low cost, providing stable service to customers and strong support for enterprise operation optimization.

[0004] However, it still has some shortcomings in actual use. First, its semantic understanding ability is limited. Existing intelligent interactive customer service systems are usually based on rule engines and keyword matching technology, which can only identify pre-set specific patterns and keywords, making it difficult to accurately understand the user's true intentions, resulting in low matching degree of the reply content. Second, the ability to handle complex problems is insufficient. The existing intelligent interactive customer service system has relatively fixed dialogue logic, lacks flexibility and contextual understanding ability, and only processes according to preset processes and rules. When the user's problem exceeds the preset range or has many steps, it is difficult to effectively handle customer problems, and it cannot adjust the processing logic according to the real-time changes of the problem. Third, the system lacks personalized service capabilities. Existing intelligent interactive customer service systems lack the ability to accumulate and call up user profiles, and cannot distinguish user identities and needs preferences, resulting in a lack of targeted services. Moreover, existing systems mostly use fixed script templates, which cannot adjust service strategies according to the context of the scenario, and cannot provide personalized knowledge recommendations. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide an intelligent interactive customer service system based on an AI large model. Through the deep collaboration between the multimodal input parsing module and the large model semantic understanding module, and through the synergistic effect of the intent preliminary detection module, the dynamic knowledge base association module, and the dialogue strategy generation module, it possesses powerful capabilities for handling complex problems and effectively solves the problems of limited semantic understanding capabilities, insufficient capabilities for handling complex problems, and lack of personalized service capabilities mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent interactive customer service system based on an AI large-scale model, comprising a multimodal input parsing module, a preliminary intent detection module, a large-scale model semantic understanding module, a dynamic knowledge base association module, a dialogue strategy generation module, a natural language generation module, and a service feedback ticket module. Multimodal input parsing module: includes an input unit and a parsing unit, wherein the input unit is used to receive multimodal user interaction requirement information uploaded by the user, and the parsing unit is used to obtain structured multimodal user interaction requirement information; The initial intent detection module includes an information filtering unit and an intent anchoring unit. The information filtering unit is used to retain valid information from the structured multimodal user interaction requirements information, and the intent anchoring unit is used to output the user's initial intent data. Large Model Semantic Understanding Module: Based on the conversational AI large model, it performs deep semantic analysis on the user's initial intent data and outputs semantic analysis results, including semantic vector representation, accurate intent classification results, entity and attribute extraction results, and contextual semantic association information; Dynamic knowledge base association module: Constructs a dynamic knowledge base, calculates the user demand matching degree based on semantic analysis results and the dynamic knowledge base, and outputs the user demand matching results, including a set of candidate knowledge items, the best knowledge recommendation results, and supplementary contextual information; Dialogue strategy generation module: Dynamically formulates user interaction strategies based on semantic analysis results and user demand matching results, and outputs a set of structured dialogue execution instructions, including interaction target instructions, dialogue flow instructions, speech style content instructions, and feedback optimization instructions; Natural Language Generation Module: Receives semantic analysis results, user demand matching results, and structured dialogue execution instruction sets, transforms them into generated prompts, outputs personalized response text through a large conversational AI model, and supports multimodal response output; Service Feedback Ticket Module: Automatically generates natural language feedback queries, receives user feedback information, performs performance evaluation based on user feedback information, outputs ticket assignment instructions, and supports simultaneous multi-channel push of ticket progress information.

[0007] The technical effects and advantages of this invention are as follows: 1. This invention, through the deep collaboration between the multimodal input parsing module and the large model semantic understanding module, comprehensively receives various forms of interactive request information uploaded by users. The large model semantic understanding module, relying on the powerful capabilities of the conversational AI large model, performs deep semantic analysis on the preliminary user intent data after initial processing, accurately extracts entity and attribute information, and mines contextual semantic association information, greatly improving the response matching degree. 2. This invention possesses powerful capabilities for handling complex problems through the synergistic effect of the preliminary intent detection module, the dynamic knowledge base association module, and the dialogue strategy generation module. The preliminary intent detection module filters out effective information from user demand information and outputs preliminary user intent data, helping the system quickly grasp the core direction of the problem. The dynamic knowledge base association module matches and calculates the semantic analysis results output by the large model with the constructed dynamic knowledge base, providing rich knowledge support for solving complex problems and improving the ability to handle complex problems. 3. This invention combines the contextual semantic association information mined by the large model semantic understanding module with the user data accumulated by the system. Based on user profiles and semantic analysis results, the dynamic knowledge base association module recommends personalized knowledge content to different users. The natural language generation module combines semantic analysis results, user demand matching results and structured dialogue execution instruction set to adjust personalized text according to user profiles and supports multimodal output, thereby improving personalized service capabilities. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0009] Figure 2 This is a schematic diagram of the method steps of the present invention.

[0010] Figure 3 This is a schematic diagram illustrating the steps for obtaining the semantic analysis results of the present invention. Detailed Implementation

[0011] 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 only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] As attached Figure 1 The intelligent interactive customer service system based on an AI big model shown includes a multimodal input parsing module, an initial intent detection module, a big model semantic understanding module, a dynamic knowledge base association module, a dialogue strategy generation module, a natural language generation module, and a service feedback ticket module.

[0013] Specifically, the output of the multimodal input parsing module is connected to the input of the preliminary intent detection module; the output of the preliminary intent detection module is connected to the input of the large model semantic understanding module; the output of the large model semantic understanding module is connected to the input of the dynamic knowledge base association module; the output of the dynamic knowledge base association module is connected to the input of the dialogue strategy generation module; the output of the dialogue strategy generation module is connected to the input of the natural language generation module; and the output of the natural language generation module is connected to the input of the service feedback ticket module.

[0014] The specific embodiments of the present invention include the following:

[0015] Multimodal input parsing module: includes an input unit and a parsing unit, wherein the input unit is used to receive multimodal user interaction requirement information uploaded by the user, and the parsing unit is used to obtain structured multimodal user interaction requirement information;

[0016] Furthermore, multimodal user interaction requirements include text input, voice input, visual input, and interaction context input.

[0017] In this embodiment, it should be specifically noted that text input information includes real-time dialogue text, form / questionnaire text, document / attachment text, and text in a specific format; voice input information includes natural voice commands / requests, auxiliary information in voice, and intermediate results of voice-to-text conversion; visual input information includes image information containing text, object / scene image information, and image metadata information; and interaction context input information includes historical interaction context, user identity / scene context, and system preset context. The system preset context refers to the basic information related to the user built into the customer service system, including but not limited to the order list bound to the user account and the user's location.

[0018] Specifically, the parsing unit receives multimodal user interaction requirement information from the input unit. For both voice and visual input information, it performs speech-to-text and image-to-text extraction, outputting textual semantics in a unified format. Based on the parsed textual semantic content, it extracts structured parameters from the user's requirements using a rule engine and regular expressions, including but not limited to order number, product type, time, and amount, outputting key parameters in key-value pair format. Based on the textual semantics, key parameters, and original modal identifiers, it associates the multimodal information of the same user with "session ID + timestamp" to generate a structured data packet, i.e., structured multimodal user interaction requirement information.

[0019] The initial intent detection module includes an information filtering unit and an intent anchoring unit. The information filtering unit is used to retain valid information from the structured multimodal user interaction requirements information, and the intent anchoring unit is used to output the user's initial intent data. In this embodiment, it should be specifically explained that the information filtering unit identifies invalid information in the structured multimodal user interaction requirement information through rule matching and lightweight semantic judgment, including garbled / meaningless characters, test statements, and duplicate submission information, retains the content containing requirement keywords or key parameters in the structured multimodal user interaction requirement information, and outputs valid information.

[0020] Furthermore, the acquisition of user preliminary intent data requires the intent anchoring unit to predefine the core intent categories of the customer service scenario, perform keyword matching and semantic similarity calculation, and output user preliminary intent data, including effective information, preliminary intent tags, matching confidence, and associated key parameters.

[0021] In this embodiment, it is important to explain that the core intent classification is pre-defined based on the user's core needs and business processing logic, and is divided into two dimensions, including primary intent classification and secondary intent classification. Primary intents include order-related, after-sales-related, logistics-related, product-related, and membership-related. Secondary intents include, but are not limited to, order inquiry, order modification, refund application, logistics inquiry, product consultation, and membership rights consultation. Each secondary intent is assigned a unique code and associated with intent description, typical keyword library, and key business parameters to form an intent classification dictionary.

[0022] It should be specifically noted that keyword matching needs to be dynamically supplemented based on the intent classification dictionary and historical user-uploaded data to form a dynamic keyword library. The valid information and the dynamic keyword library are matched precisely and fuzzily. The intents that hit the keywords are taken as candidate intents, and the number of times each candidate intent is matched with the keywords is counted.

[0023] Semantic similarity calculation specifically involves inputting effective information and a standard description of each candidate intent into a lightweight pre-trained semantic model to generate text semantic vectors. The semantic similarity between the effective information and each candidate intent is calculated using the cosine similarity formula. The candidate intent with the highest semantic similarity is selected as the initial intent label, and the semantic similarity value of the candidate intent is used as the matching confidence.

[0024] Obtaining key parameters requires using a combination of regular expressions and named entity recognition to extract business parameters that are relevant to the initial intent from the available information.

[0025] Large Model Semantic Understanding Module: Based on the conversational AI large model, it performs deep semantic analysis on the user's initial intent data and outputs semantic analysis results, including semantic vector representation, accurate intent classification results, entity and attribute extraction results, and contextual semantic association information; Furthermore, such as Figure 3 As shown, the steps for obtaining semantic analysis results are as follows: S1.1: Convert the received initial user intent data into standardized text, convert the standardized text into a token sequence and initial word vectors that the model can compute, and output the initial word vector matrix; In this embodiment, it is necessary to specifically explain that the effective content in the user's initial intent data is filtered by regular expressions, and colloquial expressions are replaced based on a preset thesaurus to obtain standardized text. The text is segmented into sub-word tokens using the model Tokenizer, and each token is mapped to a unique ID in the model dictionary to form an ID sequence. The ID sequence is then transformed into an initial word vector of fixed dimensions through the model Embedding layer to obtain an initial word vector matrix.

[0026] S1.2: Import the initial word vector matrix into the large conversational AI model, inject positional encoding, identify the relationship between tokens within the text, and output the global semantic encoding matrix; In this embodiment, it should be specifically explained that the injected position encoding specifically refers to calculating the position encoding vector of each token using the sine and cosine formulas, adding the position encoding vector to the initial word vector matrix element by element to obtain a word vector matrix with temporal information, obtaining a self-attention feature matrix through linear projection, calculating attention scores, and multi-head merging, and then performing nonlinear transformations on the self-attention feature matrix to obtain the global semantic encoding matrix.

[0027] S1.3: Generate semantic vector representations based on the global semantic encoding matrix, calculate accurate intent classification, extract entities and attributes, calculate semantic relationships between contexts, and output semantic analysis results.

[0028] In this embodiment, it should be specifically explained that semantic vector representation refers to extracting the semantic vector of the entire sentence from the global semantic encoding matrix to obtain the text semantic vector; calculating accurate intent classification refers to determining the probability that the user input belongs to a preset sub-intent, taking the text semantic vector and performing linear mapping through the classification weight matrix to generate the original score, converting the original score into the probability of each intent, selecting the intent with the highest probability as the accurate intent label, using the corresponding probability as the confidence level, and outputting the accurate intent classification result; extracting entities and attributes refers to generating the entity label score for each token by passing the global semantic encoding matrix through the NER weight matrix, taking the label with the highest probability after Softmax, performing entity merging and attribute association, calculating the attention weight between the entity token and the attribute description token, and if the weight is >0.7, it is associated as an attribute, and outputting the entity and attribute extraction result; calculating the context semantic association involves generating a historical semantic encoding matrix from the historical text of multi-turn dialogues, calculating the cosine similarity between the current and historical semantic encoding matrices, and outputting the context semantic association information.

[0029] Dynamic knowledge base association module: Constructs a dynamic knowledge base, calculates the user demand matching degree based on semantic analysis results and the dynamic knowledge base, and outputs the user demand matching results, including a set of candidate knowledge items, the best knowledge recommendation results, and supplementary contextual information; Furthermore, the dynamic knowledge base supports importing data from multiple sources and can automatically update knowledge through scheduled crawling and API interface synchronization. The knowledge in the dynamic knowledge base is stored in the form of "semantic vector + text description".

[0030] In this embodiment, it should be specifically noted that the import of multi-source data includes, but is not limited to, updates to corporate website policy documents, product manuals, and adjustments to activity rules. The dynamic knowledge base includes a four-layer architecture: data layer, structure layer, index layer, and update layer. The dynamic knowledge base establishes relationships between knowledge by constructing a knowledge graph and builds multi-dimensional indexes, including semantic vector indexes, attribute indexes, and association indexes.

[0031] Furthermore, the calculation steps for user demand matching degree are as follows: S2.1: Based on the semantic analysis results and dynamic knowledge base, obtain user demand analysis parameters, including user demand semantic vector, knowledge base semantic vector, user demand entity type set, knowledge-related entity type set, user demand time, and knowledge effective time; In this embodiment, it is necessary to specifically explain that the user demand semantic vector is obtained through the user's initial intent data, the knowledge semantic vector is obtained through the dynamic knowledge base, the user demand entity type set and user demand time are obtained based on the semantic analysis results output by the large model semantic understanding module, and the knowledge associated entity type set and knowledge effective time are obtained through the dynamic knowledge base. The knowledge effective time refers to the "effective start time" and "effective end time" in the metadata of the knowledge entry in the dynamic knowledge base.

[0032] S2.2: Based on user requirement analysis parameters, the semantic vector matching degree S and entity type overlap degree M are calculated. t Entity attribute matching degree M a Timeliness adaptability (F), contextual relevance (R), and knowledge confidence (C); In this embodiment, it should be specifically noted that the semantic vector matching degree S is calculated using the cosine similarity formula to determine the spatial correlation between the user's demand semantic vector and the knowledge base semantic vector, retaining knowledge with a semantic vector matching degree ≥ 0.6; the entity type overlap degree M... t It needs to be based on the user's requirement entity type set E u The set of entity types E associated with knowledge d Through the formula: ,

[0033] The calculation yields a value where Num represents the number of geometric elements, and M is retained. t Knowledge with a value ≥0.5; Obtaining the entity attribute matching degree F requires extracting the key attributes of the user's requirement entity, extracting the usage attribute constraints of the knowledge base, and calculating it through the attribute weighting formula, retaining M. a Knowledge level ≥ 0.6; Timeliness adaptation is calculated based on user demand time T. u and the effective time range of knowledge [T] ds T de ], through the formula: ,

[0034] The calculation yields ΔT. th For a time threshold, such as 30 days, if the time exceeds 30 days, F=0, indicating the knowledge is expired; knowledge with F≥0.7 is retained. Contextual relevance R is calculated by obtaining historical global semantic vectors and using the cosine similarity formula to determine the relevance between the historical global semantic vectors and the knowledge semantic vectors; knowledge with R≥0.7 is retained. Knowledge confidence C is obtained by extracting the manual verification status V and the last update day O from the knowledge base metadata, using the formula: ,

[0035] The calculation shows that ΔTmax is the maximum allowable number of days of delay, which can be 180 days. If it exceeds this, this term is 0, and we retain the knowledge that C≥0.8.

[0036] S2.3: Define the compliance dimensions and key dimensions, where the key dimensions are semantic vector matching degree, entity attribute matching degree, and timeliness adaptability, expressed by the formula: ,

[0037] The user demand matching degree SI is calculated, where K represents the number of dimensions that meet the criteria, and K / 6 is the dimension compliance rate.

[0038] In this embodiment, it should be specifically explained that the compliance dimension refers to the number of dimensions that meet the screening rules. There are a total of 6 dimensions. The compliance rate of each dimension reflects the overall adaptability of each dimension, the average score of the key dimensions reflects the core adaptability, and the user demand matching degree ranges from 0 to 1. The higher the value, the higher the overall user demand matching degree.

[0039] It should be specifically noted that the candidate knowledge item set is initially screened based on user demand matching degree, retaining knowledge items with user demand matching degree ≥ 0.7, and ensuring that the candidate knowledge item set meets the following conditions: semantic vector matching degree ≥ 0.6 + entity attribute matching degree ≥ 0.6 + timeliness adaptability ≥ 0.8. The items are then sorted in descending order based on comprehensive matching degree, knowledge confidence degree, and update time, and the basic information, matching degree data, and association identifier are output. The optimal knowledge recommendation result needs to be obtained based on the candidate knowledge item set through multi-dimensional weighted decision-making and business rule verification. The user demand matching degree, entity attribute matching degree, contextual relevance degree and knowledge priority are weighted and scored, with weights of 0.4, 0.3, 0.2 and 0.1 respectively. The knowledge with the highest score is selected as the optimal result. If there are exceptions to the optimal knowledge, it needs to be further compared with the user entity attributes to ensure that there are no conflicts. The complete knowledge content, recommendation basis and confidence level are output. The contextual supplementary information needs to clarify the matching relationship between knowledge and entities in user needs, extract the core semantic focus of user needs and knowledge, explain the scenario adaptation, recommend 1-2 pieces of knowledge that the user may need later, and output the related logic, extended knowledge and supplementary prompts.

[0040] Dialogue strategy generation module: Dynamically formulates user interaction strategies based on semantic analysis results and user demand matching results, and outputs a set of structured dialogue execution instructions, including interaction target instructions, dialogue flow instructions, speech style content instructions, and feedback optimization instructions; Furthermore, the interaction target instructions include target type, target description, target priority, and completion criteria; the dialogue flow instructions include flow nodes, exception handling rules, and flow termination conditions; the speech style and content instructions include speech style, core content elements, format requirements, and entity embedding requirements; and the feedback optimization instructions include feedback listening type, optimization strategy, optimization priority, and strategy update mechanism.

[0041] In this embodiment, it should be specifically noted that the interaction target instruction needs to be obtained based on the accurate intent classification result and the optimal knowledge recommendation result. The target type includes, but is not limited to, knowledge answering, intent clarification, and parameter supplementation. The target description refers to the specific action and content to be achieved. The dialogue flow instruction needs to be obtained based on the interaction target instruction and contextual semantic association information. The flow node refers to the action node arranged according to the execution order. The speech style content instruction needs to be obtained based on the optimal knowledge recommendation result and entity and attribute extraction result. The core content elements refer to the knowledge points that must be included and the prohibited content. The entity embedding requirement refers to how to naturally integrate user entity information into the speech. The feedback optimization instruction needs to be obtained based on the candidate knowledge item set and intent classification confidence. The feedback listening type refers to the user feedback scenarios that need to be paid attention to. The strategy update mechanism refers to the strategy parameters that are periodically optimized based on feedback data.

[0042] Natural Language Generation Module: Receives semantic analysis results, user demand matching results, and structured dialogue execution instruction sets, transforms them into generated prompts, outputs personalized response text through a large conversational AI model, and supports multimodal response output; Furthermore, the generated prompts include a description of the requirements, supporting knowledge, and contextual parameters; Multimodal response output includes voice response output, text and image response output, and interactive response output.

[0043] In this embodiment, it should be specifically explained that the core demand is based on the accurate intent classification results and is used to clarify the user's core demand direction; the knowledge support includes core knowledge content, knowledge association basis and knowledge boundary restrictions. The core knowledge content comes from the best knowledge recommendation results, the knowledge association basis comes from contextual supplementary information and is used to explain the reason for the matching of knowledge and demand, and the context parameters come from the candidate knowledge item set and knowledge metadata and are used to clarify unusable knowledge.

[0044] It should be further explained that the output of personalized response text needs to be connected to an external image system to receive user profiles, and to perform in-depth optimization based on user profile data and historical interaction feedback data. The personalized response text is optimized based on the user profile data, including age, membership level, historical interaction style, and preferred language. Based on historical interaction feedback data, the user's feedback on the response in the past interaction is retrieved to adjust the current response style.

[0045] Service Feedback Ticket Module: Automatically generates natural language feedback queries, receives user feedback information, performs performance evaluation based on user feedback information, outputs ticket assignment instructions, and supports simultaneous multi-channel push of ticket progress information.

[0046] Furthermore, the work order assignment instructions include basic work order information, assignment targets, execution requirements, and associated data.

[0047] In this embodiment, it is important to explain that the performance evaluation is based on user feedback information. A three-dimensional hierarchical evaluation rule is constructed. Through keyword extraction and rule matching, four scenarios are distinguished: clear satisfaction, partial resolution / vague feedback, clear dissatisfaction, and emergency / complaint. Keywords in user feedback information are extracted, and the parsed feedback tags are matched with preset rules to determine whether to generate a work order and the type of work order. The conclusion of no work order required / needs a regular work order / needs an emergency work order is output.

[0048] It should be noted that work order allocation needs to generate work order allocation instructions based on work order priority, demand type, user level, and agent resource status. Emergency work orders need to be marked with an emergency identifier and escalation mechanism on the basis of ordinary work orders. The corresponding agent group should be matched according to the demand type, agents should be selected according to preset priority rules, and the work order allocation instructions should be sent to the agent workbench of the target agent.

[0049] like Figure 2 As shown, this invention provides an intelligent interactive customer service method based on a large AI model, with the following steps: S1: Receive multimodal user interaction requirement information uploaded by users through the multimodal input parsing module, and obtain structured multimodal user interaction requirement information; S2: The preliminary intent detection module retains valid information from the structured multimodal user interaction needs information and outputs preliminary user intent data; S3: The large-scale model semantic understanding module performs deep semantic analysis on the user's initial intent data based on the conversational AI large-scale model and outputs the semantic analysis results. S4: Calculate the user demand matching degree based on semantic analysis results and dynamic knowledge base through the dynamic knowledge base association module, and output the user demand matching result; S5: The dialogue strategy generation module dynamically formulates user interaction strategies based on semantic analysis results and user demand matching results, and outputs a set of structured dialogue execution instructions; S6: Receives and outputs semantic analysis results, user demand matching results, and structured dialogue execution instruction sets through the natural language generation module, and outputs personalized response text; S7: Automatically generates natural language feedback queries through the service feedback work order module, receives user feedback information, performs performance evaluation based on user feedback information, and outputs work order assignment instructions.

[0050] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. In conclusion, 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 protection scope of the present invention.

Claims

1. An intelligent interactive customer service system based on an AI large-scale model, characterized in that: It includes a multimodal input parsing module, a preliminary intent detection module, a large-scale model semantic understanding module, a dynamic knowledge base association module, a dialogue strategy generation module, a natural language generation module, and a service feedback ticket module. The multimodal input parsing module includes an input unit and a parsing unit, wherein the input unit is used to receive multimodal user interaction requirement information uploaded by the user, and the parsing unit is used to obtain structured multimodal user interaction requirement information. The preliminary intent detection module includes an information filtering unit and an intent anchoring unit, wherein the information filtering unit is used to retain valid information in the structured multimodal user interaction requirement information, and the intent anchoring unit is used to output preliminary user intent data. The large model semantic understanding module performs deep semantic analysis on the user's initial intent data based on the conversational AI large model, and outputs semantic analysis results, including semantic vector representation, accurate intent classification results, entity and attribute extraction results, and contextual semantic association information. The dynamic knowledge base association module constructs a dynamic knowledge base, calculates the user demand matching degree based on semantic analysis results and the dynamic knowledge base, and outputs the user demand matching results, including a set of candidate knowledge items, the optimal knowledge recommendation results, and supplementary contextual information. The dialogue strategy generation module dynamically formulates user interaction strategies based on semantic analysis results and user demand matching results, and outputs a set of structured dialogue execution instructions, including interaction target instructions, dialogue flow instructions, speech style content instructions, and feedback optimization instructions. The natural language generation module receives and outputs semantic analysis results, user demand matching results, and structured dialogue execution instruction sets, transforms them into generation prompts, outputs personalized response text through a conversational AI big model, and supports multimodal response output. The service feedback work order module automatically generates natural language feedback queries, receives user feedback information, judges the effect based on user feedback information, outputs work order allocation instructions, and supports synchronous multi-channel push of work order progress information.

2. The intelligent interactive customer service system based on an AI large model according to claim 1, characterized in that: The multimodal user interaction requirements information includes text input information, voice input information, visual input information, and interaction context input information.

3. The intelligent interactive customer service system based on an AI large model according to claim 1, characterized in that: The acquisition of the user's initial intent data requires the intent anchoring unit to predefine the core intent classification of the customer service scenario, perform keyword matching and semantic similarity calculation, and output the user's initial intent data, including valid information, initial intent tags, matching confidence, and associated key parameters.

4. The intelligent interactive customer service system based on an AI large model according to claim 1, characterized in that: The steps for obtaining the semantic analysis results are as follows: S1.1: Convert the received initial user intent data into standardized text, convert the standardized text into a token sequence and initial word vectors that the model can compute, and output the initial word vector matrix; S1.2: Import the initial word vector matrix into the large conversational AI model, inject positional encoding, identify the relationship between tokens within the text, and output the global semantic encoding matrix; S1.3: Generate semantic vector representations based on the global semantic encoding matrix, calculate accurate intent classification, extract entities and attributes, calculate semantic relationships between contexts, and output semantic analysis results.

5. The intelligent interactive customer service system based on an AI large model according to claim 1, characterized in that: The dynamic knowledge base supports importing data from multiple sources and can automatically update knowledge through scheduled crawling and API interface synchronization. The knowledge in the dynamic knowledge base is stored in the form of "semantic vector + text description".

6. The intelligent interactive customer service system based on an AI large model according to claim 1, characterized in that: The steps for calculating the user demand matching degree are as follows: S2.1: Based on the semantic analysis results and dynamic knowledge base, obtain user demand analysis parameters, including user demand semantic vector, knowledge base semantic vector, user demand entity type set, knowledge-related entity type set, user demand time, and knowledge effective time; S2.2: Based on user requirement analysis parameters, the semantic vector matching degree S and entity type overlap degree M are calculated. t Entity attribute matching degree M a Timeliness adaptability (F), contextual relevance (R), and knowledge confidence (C); S2.3: Define the compliance dimensions and key dimensions, where the key dimensions are semantic vector matching degree, entity attribute matching degree, and timeliness adaptability, expressed by the formula: ， The user demand matching degree SI is calculated, where K represents the number of dimensions that meet the criteria, and K / 6 is the dimension compliance rate.

7. The intelligent interactive customer service system based on an AI large model according to claim 1, characterized in that: The interactive target instructions include target type, target description, target priority, and completion criteria; the dialogue flow instructions include flow nodes, exception handling rules, and flow termination conditions; the speech style and content instructions include speech style, core content elements, format requirements, and entity embedding requirements. Feedback optimization instructions include feedback listening type, optimization strategy, optimization priority, and strategy update mechanism.

8. The intelligent interactive customer service system based on an AI large model according to claim 1, characterized in that: The generated prompts include a requirement description, knowledge support, and context parameters; Multimodal response output includes voice response output, text and image response output, and interactive response output.

9. The intelligent interactive customer service system based on an AI large model according to claim 1, characterized in that: The work order allocation instruction includes basic work order information, allocation target, execution requirements, and associated data.

Images