Multi-task guide type dialogue processing method and system for intelligent customer service
Through deep semantic understanding and business meta-model-driven state flow management, the intelligent customer service system can dynamically build query targets and conditions, solving the efficiency and experience problems of existing systems in multi-task dialogue processing, and achieving higher task processing accuracy and user satisfaction.
Patent Information
- Application Number
- CN202410947390.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-16
- Publication Date
- 2024-11-12
AI Technical Summary
The existing intelligent customer service system lacks the ability to dynamically build and adjust query targets and conditions when handling complex multi-task conversations, and fails to effectively manage and control state transitions, resulting in poor processing efficiency and user experience.
A multi-task wizard dialogue processing method is adopted to analyze user input through deep semantic understanding, determine query targets and conditions, build a function call chain, and use the business processing meta-model to drive state flow, and dynamically call business functions to deal with different task scenarios.
It significantly improves task processing accuracy and flexibility, ensures smooth and logically coherent dialogue processes in multi-task scenarios, and improves user interaction experience and overall service quality.
Smart Images

Figure CN118941295A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent customer service, and in particular to a multi-task wizard-style dialogue processing method and system for intelligent customer service. Background Art
[0002] In the current intelligent customer service field, the research and development of dialogue processing systems has made significant progress, especially in natural language understanding and task-oriented dialogue. Existing mainstream systems usually use general large models or natural language processing technology to parse user input information, and perform single or limited-range task processing by identifying keywords, topics or intentions. For example, based on semantic analysis and syntactic structure parsing technology, some systems can achieve preliminary division of the subject, predicate, object, attributive, adverbial, complement and other components of user sentences to understand the basic needs of users.
[0003] However, there are several limitations in the existing technology. Traditional intelligent customer service dialogue systems are often unable to cope with complex multi-task scenarios. Most of them can only respond according to preset processes and fixed templates, lack the ability to dynamically build and adjust query targets and conditions, and lack the mechanism to flexibly call different processing functions according to different task scenarios. In addition, in the process of handling multi-task dialogue flow, there is no effective solution based on meta-model for efficient management and control of state transitions. Summary of the invention
[0004] In view of the above-mentioned deficiencies in the prior art, an object of the present invention is to propose a multi-task wizard dialogue processing method and system for intelligent customer service, which can at least partially solve the problems existing in the above-mentioned background technology.
[0005] In order to achieve the above object, the technical solution adopted by the present invention is:
[0006] A multi-task wizard-style dialogue processing method for intelligent customer service, comprising:
[0007] S11: parsing a sentence input by a user, wherein the parsing includes deep semantic understanding;
[0008] S12: Determine the query target and query conditions according to the parsing result;
[0009] S13: According to the query target and query conditions, a function call chain is constructed, and the state flow is driven by combining the business processing metamodel, and the corresponding business function is called and executed; wherein the business processing metamodel includes all business states and the transfer logic between business states.
[0010] Another technical solution adopted by the present invention is:
[0011] A multi-task wizard-style dialogue processing system for intelligent customer service, comprising:
[0012] A syntax analysis module, used to parse the sentences input by the user, wherein the parsing includes deep semantic understanding;
[0013] The target and condition identification module is used to determine the query target and query condition according to the parsing result;
[0014] A dynamic task mobilization module is used to build a function call chain according to the query target and query conditions, and call and execute the corresponding business function according to the function call chain;
[0015] The metamodel-driven state flow management module is used to control the business state flow by using the business processing metamodel; the business processing metamodel includes all business states and the transfer logic between business states.
[0016] The beneficial effects of the present invention are as follows: sentence parsing includes deep semantic understanding, which can more accurately capture and understand the user's true intentions, and on this basis dynamically construct query targets and conditions closely related to user needs, thereby significantly improving the system's task processing accuracy and flexibility. Multi-task business functions can be dynamically called according to query targets and conditions, so that corresponding task operations can be performed in different scenarios to meet users' diverse service needs. The business metamodel effectively manages state transfers during the conversation process, ensuring that the conversation process in multi-task scenarios is smooth and logically coherent, improving user interaction experience, and further promoting the technological innovation and development of intelligent customer service systems. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 The figure is a flow chart of a multi-task wizard-style dialogue processing method for intelligent customer service according to an embodiment of the present invention;
[0018] Figure 2 Shown is a schematic diagram of the framework of a multi-task wizard dialogue processing system for intelligent customer service according to an embodiment of the present invention. DETAILED DESCRIPTION
[0019] In order to more clearly understand the technical content, achieved purposes and effects of the present invention, the present invention is described in detail below in conjunction with specific implementation methods and in conjunction with the accompanying drawings. It should be noted that, in the absence of conflict, the implementation methods of the present invention and the features in the implementation methods can be combined with each other. In the following description, many specific details are set forth in order to fully understand the present invention. The described implementation methods are only part of the implementation methods of the present invention, not all implementation methods. Based on the implementation methods in the present invention, all other implementation methods obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present invention.
[0020] Please refer to Figure 1As shown, the technical solution provided by the present invention is:
[0021] A multi-task wizard-style dialogue processing method for intelligent customer service, comprising:
[0022] S11: parsing a sentence input by a user, wherein the parsing includes deep semantic understanding;
[0023] S12: Determine the query target and query conditions according to the parsing result;
[0024] S13: According to the query target and query conditions, a function call chain is constructed, and the state flow is driven by combining the business processing metamodel, and the corresponding business function is called and executed; wherein the business processing metamodel includes all business states and the transfer logic between business states.
[0025] From the above description, it can be seen that the beneficial effect of the multi-task wizard dialogue processing method for intelligent customer service of the present invention is that through sentence parsing and target and condition identification including deep semantic understanding, it can more accurately grasp the user's intention, effectively reduce misunderstanding and misjudgment rate, and improve user satisfaction. By constructing a function call chain according to the query target and conditions, and dynamically calling the business function, it is possible to flexibly adjust the processing strategy according to the actual business scenario, and easily cope with the ever-changing user needs and market environment. The state flow management method driven by the business processing metamodel greatly improves the coordination and fluency of each link in the dialogue process, reduces the possibility of invalid dialogue and repeated interaction, and thus improves the overall dialogue processing efficiency and service quality.
[0026] Step S11 specifically includes: deep mining of the sentences input by the user, using a predefined knowledge base and a set of complex sentence rules customized for the characteristics of Chinese to effectively identify the multi-level, nested subject, predicate, object, attributive, adverbial, and complement grammatical structures in the question. This process is not limited to the basic syntactic level division, but also includes deep semantic understanding that is sensitive to the context to ensure that the root of the user's intention can be accurately captured. For example, when faced with complex sentences containing multiple clauses or long-distance dependencies, it can effectively reveal their internal relevance and provide accurate syntactic analysis results for subsequent processing.
[0027] In an optional implementation manner, step S11 specifically includes:
[0028] S111: In a preset service scenario, the user's potential query targets are classified, and corresponding basic semantic units are set for each type of query target; these units, as basic modules for expressing intentions at the semantic level, can accurately capture the essence of user needs.
[0029] S112: Use a pre-trained neural network model to fine-tune the results of S111 to obtain key semantic slot content; this process ensures the system's deep understanding and effective use of user input information.
[0030] S113: Systematically sort out the knowledge points of the semantic slot content, and add keywords and subject terms according to the preset knowledge base; among them, subject terms are defined and published by authoritative organizations, usually covering proper nouns, noun phrases and their variants. Searching through subject terms can effectively reduce missed detection and false detection, and improve the accuracy of information retrieval. Keywords are derived from expert knowledge bases or prompt projects, which highly summarize the topic of the article and facilitate targeted search of content with clear topic orientation.
[0031] S114: Create a multi-dimensional and multi-level index structure for the same knowledge point; multi-level indexing enables the same knowledge point to still respond efficiently in multi-attribute query scenarios, allowing the system to quickly locate and extract specific attribute data without traversing all knowledge content, significantly improving knowledge query efficiency.
[0032] S115: Convert the preset knowledge base into a structured knowledge graph, and build a triple model with entities, attributes, and relationships as the core; entities represent objects or phenomena in the real world, attributes describe their characteristics and states, and relationships reveal the connections between entities. This knowledge graph processing method enhances the organization, comprehensibility, and reasonability of knowledge, and provides strong support for the system to answer complex questions and conduct in-depth conversations.
[0033] S116: According to the index structure, the triple model is used to retrieve the semantic slot content of step S112 to obtain corresponding parsed content.
[0034] After completing the in-depth analysis of the user's input sentence, we apply the principles of Chinese logical semantics to conduct detailed reasoning and analysis on the subject, predicate, object, attributive, adverbial, complement and other elements in the syntactic structure to clarify the dialogue goals and extract query conditions.
[0035] In an optional implementation, the parsed result includes one or more elements of a subject, a predicate, an object, an attributive, an adverbial, and a complement; step S12 specifically includes:
[0036] S121: Extract noun phrases from the parsed results; the present invention focuses on identifying subjects and objects, and these two core noun phrases are usually composed of nouns and their qualifiers or modifiers. It is worth noting that knowledge points are not organized according to events, but are built around entities, especially noun phrases. Therefore, the effective extraction of noun phrases is the key to determining the query target.
[0037] S122: Extract predicates from the parsed results, identify events based on the extracted predicates and noun phrases; further parse the predicate part to reveal the event relationship between noun phrases, such as evaluation, comparison, query, filtering, statistics, etc. The existence of these dynamic calculation and processing requirements means that even in the absence of predicate information, complete knowledge can be captured by the combination of noun phrases alone.
[0038] S123: analyzing the parallel and limiting relationships between noun phrases, performing cascade processing on limiting relationships, and performing cross multiplication operations on parallel relationships;
[0039] S124: construct a grammatical component tree according to the results of steps S121-S123; perform a search according to the grammatical component tree to generate a query target and a query condition.
[0040] In an optional implementation manner, step S124 specifically includes:
[0041] S1241: construct a grammatical component tree according to the results of steps S121-S123; establish the subject and object as the query target through minimum recursive traversal;
[0042] S1242: Perform cross-product processing on related adverbials and complements to generate detailed query conditions.
[0043] These query targets and query conditions are sent to the knowledge base for query retrieval in the subsequent process.
[0044] In an optional implementation, custom business rules are introduced to divide query conditions and targets into "precise mode" and "fuzzy mode". The precise mode is suitable for scenarios where users query fixed and clear information, and information matching is strictly performed at this time; when user needs are vague or seeking related recommendations, it switches to fuzzy mode and uses advanced algorithms such as similarity calculation and semantic association for flexible matching and related recommendations.
[0045] Through the above steps, the target object and query conditions are accurately identified, which significantly enhances the adaptability and interaction accuracy of intelligent customer service. No matter how user needs change, it can respond calmly and provide highly compatible diversified services, fully demonstrating the powerful functions and intelligent level of multi-task wizard dialogue processing.
[0046] On the basis of clarifying the query objectives and query conditions, through the carefully designed business processing metamodel intelligent scheduling mechanism, the corresponding processing strategies are accurately called to deal with multi-task wizard dialogues.
[0047] In an optional implementation manner, step S13 specifically includes:
[0048] S131: searching for a query target function from a preset function library according to the determined query target. If the query target function is not found, triggering a preset fallback function; if the query target function is found, executing the following steps;
[0049] S132: Decomposing the query target and the query condition and mapping them to a function call chain;
[0050] S133: Using the business processing metamodel to drive state transfer, calling and executing the corresponding business function according to the function call chain. A user statement may contain multiple goals and conditions. By using regular expressions, these elements are organically organized into a logically clear function execution chain.
[0051] In an optional implementation manner, the S133 specifically includes:
[0052] The context information of the conversation is embedded in the function call chain, and the preset business processing metamodel is used for node judgment. The corresponding business call process is triggered according to the processing node where the conversation is located. The business processing metamodel plays a key navigation role here, triggering the corresponding business call process according to the conversation process, ensuring that the conversation process is closely aligned with actual business needs.
[0053] In an optional implementation, the business processing metamodel includes several key business processing nodes. When the query target or query condition includes multiple tasks, S133 specifically includes:
[0054] The disassembled query conditions and query parameters are input into the processing nodes of each business processing metamodel (such as business processing node, business cancellation node, chat node, knowledge search node, continue processing node, etc.), triggering the call of the corresponding business function and executing the called business function.
[0055] The execution results of each node are integrated in an orderly manner to form the final dialogue output, thereby achieving efficient advancement of each stage of the dialogue, accurately managing state changes, and ensuring that no matter how complex and changeable the business process is, it can quickly respond to and accurately complete various dialogue tasks, greatly improving overall efficiency and service quality.
[0056] Through this highly automated and adaptive process management method, users can enjoy a smooth and efficient conversation experience, meeting their diverse needs in various business scenarios.
[0057] The present invention also provides a multi-task wizard-style dialogue processing system for intelligent customer service, comprising:
[0058] A syntax analysis module, used to parse the sentences input by the user, wherein the parsing includes deep semantic understanding;
[0059] The target and condition identification module is used to determine the query target and query condition according to the parsing result;
[0060] A dynamic task mobilization module is used to build a function call chain according to the query target and query conditions, and call and execute the corresponding business function according to the function call chain;
[0061] The metamodel-driven state flow management module is used to control the business state flow by using the business processing metamodel; the business processing metamodel includes all business states and the transfer logic between business states.
[0062] From the above description, it can be seen that the beneficial effect of the multi-task wizard dialogue processing system for intelligent customer service of the present invention is that: by deeply analyzing the user input content, it can more accurately capture and understand the user's true intention, and on this basis dynamically construct query goals and conditions closely related to user needs, thereby significantly improving the system's task processing accuracy and flexibility. The mechanism of dynamically calling multiple task function processing logics according to goals and conditions enables the system to perform corresponding task operations in different scenarios to meet the user's diverse service needs. Multi-task state flow management based on the business processing metamodel effectively manages the state transfer during the dialogue process, ensures that the dialogue process in the multi-task scenario is smooth and logically coherent, improves the user interaction experience, and further promotes the technological innovation and development of the intelligent customer service system.
[0063] In an optional implementation manner, the dynamic task mobilization module is specifically used to:
[0064] The query target function is retrieved from the preset function library according to the determined query target. If the query target function is not retrieved, the preset fallback function is triggered; if the query target function is retrieved, the query target and query conditions are disassembled and mapped to the function call chain;
[0065] The corresponding business function is called and executed according to the function call chain.
[0066] In an optional implementation, the business processing metamodel includes several key business processing nodes, the dynamic task mobilization module embeds the context information of the conversation into the function call chain, the metamodel-driven state flow management module, the dynamic task mobilization module triggers the corresponding business call process according to the conversation node; and inputs the judged node into the processing node of each business processing metamodel, triggering the dynamic task mobilization module to call the corresponding business function, and execute the called business function.
[0067] In an optional implementation, it further includes: a function library storing various business functions, wherein the function library is dynamically linked to the dynamic task mobilization module.
[0068] Please refer to Figure 1, Embodiment 1 of the present invention is:
[0069] A multi-task wizard-style dialogue processing method for intelligent customer service, comprising:
[0070] S21: Construct a knowledge base, a function library and a business processing metamodel, wherein the knowledge base stores knowledge points related to the dialogue, and the function library stores task processing functions related to the dialogue, wherein the functions include target functions and business functions. The business processing metamodel includes all business states, the transfer logic between business states and several key business processing nodes, wherein the key business processing nodes include business processing nodes, business cancellation nodes, chat nodes, knowledge search nodes and continue processing nodes.
[0071] S22: parsing the sentence input by the user, wherein the parsing includes deep semantic understanding; S22 specifically includes:
[0072] Step S11 specifically includes:
[0073] S221: classifying potential query targets of the user in a preset service scenario, and setting corresponding basic semantic units for each type of query target;
[0074] S222: Use the pre-trained neural network model to fine-tune the results of S221 to obtain the key semantic slot content;
[0075] S223: Systematically sort out the knowledge points of the semantic slot content, and add keywords and subject terms according to the preset knowledge base;
[0076] S224: Creating a multi-dimensional and multi-level index structure for the same knowledge point;
[0077] S225: Convert the preset knowledge base into a structured knowledge graph and build a triple model with entities, attributes and relationships as the core;
[0078] S226: According to the index structure, the triple model is used to retrieve the semantic slot content of step S222 to obtain a corresponding syntactic structure, wherein the syntactic structure includes one or more elements of subject, predicate, object, attributive, adverbial, and complement.
[0079] S23: Determine the query target and query conditions according to the parsing result; S23 specifically includes:
[0080] S231: extracting noun phrases from the parsed results;
[0081] S232: extracting predicates from the parsed results, and identifying events based on the extracted predicates and noun phrases;
[0082] S233: Analyze the parallel and limiting relationships between noun phrases, perform cascade processing on limiting relationships, and perform cross multiplication operations on parallel relationships; construct a semantic expression containing one or more of the subject, predicate, object, attributive, adverbial, and complement;
[0083] S234: constructing a grammatical component tree according to the results of steps S211-S233; establishing the subject and the object as the query target through minimum recursive traversal;
[0084] S235: Perform cross-product processing on related adverbials and complements to generate query conditions.
[0085] S24: construct a function call chain according to the query target and query conditions, drive the state flow in combination with the business processing metamodel, and call and execute the corresponding business function; the business processing metamodel includes S24 specifically including:
[0086] S241: searching for a query target function from a preset function library according to the determined query target. If the query target function is not found, triggering a preset fallback function; if the query target function is found, executing the following steps;
[0087] S242: Decompose the query target and query condition and map them to a function call chain;
[0088] S243: Embed the context information of the conversation into the function call chain, and use the preset business processing metamodel to perform node judgment; trigger the corresponding business call process according to the processing node where the conversation is located; when the query target or query condition contains multiple tasks, input the disassembled query condition and query parameter into the processing node of each business processing metamodel, trigger the call of the corresponding business function, and execute the called business function.
[0089] Please refer to Figure 2 , Embodiment 2 of the present invention is:
[0090] A multi-task wizard-style dialogue processing system for intelligent customer service, corresponding to the multi-task wizard-style dialogue processing method for intelligent customer service of the above embodiment, includes:
[0091] The knowledge base is used to store knowledge points related to the dialogue.
[0092] The function library is used to store task processing functions related to the dialogue, wherein the functions include target functions and business functions. The function library is dynamically connected to the dynamic task mobilization module.
[0093] The syntactic analysis module is used to parse the sentences input by the user, and the analysis includes deep semantic understanding; specifically, the syntactic analysis module performs the following steps: S221: in the preset service scenario, the user's potential query targets are classified, and corresponding basic semantic units are set for each type of query target; S222: the pre-trained neural network model is used to fine-tune the result of S221 to obtain the key semantic slot content; S223: the knowledge points of the semantic slot content are systematically sorted out, and keywords and subject terms are added according to the preset knowledge base; S224: a multi-dimensional and multi-level index structure is created for the same knowledge points; S225: the preset knowledge base is converted into a structured knowledge graph, and a triple model is constructed with entities, attributes and relationships as the core; S226: according to the index structure, the triple model is used to retrieve the semantic slot content of step S222 to obtain the corresponding syntactic structure, and the syntactic structure includes one or more elements of subject, predicate, object, attributive, adverbial, and complement.
[0094] The target and condition identification module is used to determine the query target and query condition according to the results of the analysis; specifically, the target and condition identification module performs the following steps: S231: extract noun phrases from the results of the analysis; S232: extract predicates from the results of the analysis, and identify events based on the extracted predicates and noun phrases; S233: analyze the parallel and limiting relationships between noun phrases, perform cascade processing on the limiting relationships, and perform cross multiplication operations on the parallel relationships; construct a semantic expression containing one or more of the subject, predicate, object, attributive, adverbial, and complement; S234: construct a grammatical component tree according to the results of steps S231-S233; through the minimum number recursive traversal, establish the subject and object as the query target; S235: perform cross multiplication processing on the relevant adverbials and complements to generate query conditions.
[0095] The dynamic task mobilization module is used to retrieve the query target function from the preset function library according to the determined query target. If the query target function is not retrieved, the preset fallback function is triggered; if the query target function is retrieved, the query target and query conditions are disassembled and mapped to the function call chain; the context information of the conversation is embedded in the function call chain, and the corresponding business call process is triggered according to the processing node where the conversation is located; when processing multiple tasks, the corresponding business function is triggered and executed according to each task processing node.
[0096] The metamodel-driven state flow management module includes a business processing metamodel, which includes all business states, the transfer logic between business states, and several key business processing nodes, and the key business processing nodes include business processing nodes, business cancellation nodes, chat nodes, knowledge search nodes, and continue processing nodes. The metamodel-driven state flow management module is used to use the business processing metamodel to determine the processing node where the dialogue is located according to the query target and query conditions, and drive the dynamic task mobilization module to call the corresponding function according to the processing node.
[0097] In summary, the multi-task wizard-style dialogue processing method and system for intelligent customer service proposed in the present invention have the following advantages:
[0098] 1. Improve conversation understanding and response accuracy: Through deep syntactic analysis and intelligent target recognition, the system can more accurately grasp user intent, effectively reduce misunderstandings and misjudgment rates, and improve user satisfaction.
[0099] 2. Enhance system adaptability and flexibility: The dynamic task function calling mechanism allows the system to flexibly adjust the processing strategy according to the actual business scenario, and easily respond to the ever-changing user needs and market environment.
[0100] 3. Optimize business processes and dialogue efficiency: Metamodel-driven state flow management greatly improves the coordination and fluency of each link in the dialogue process, reduces the possibility of invalid dialogue and repeated interactions, and thus improves the overall dialogue processing efficiency and service quality.
[0101] 4. Broaden the scope of application and expand possibilities: The technical solution of the present invention provides a powerful multi-tasking processing capability for the intelligent customer service system, which can be widely used in e-commerce consulting, financial management, medical health and other fields, providing users with a more personalized and intelligent service experience.
[0102] To sum up, the beneficial effects brought about by technological innovation in the present invention have successfully solved many problems existing in the existing intelligent customer service system when dealing with complex multi-task conversations, and achieved all-round optimization and upgrading from conversation understanding to task execution to state flow, thereby promoting substantial progress in the field of intelligent customer service technology.
[0103] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Therefore, any modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention are still within the scope of the technical solution of the present invention.
Claims
1. A multi-task wizard dialogue processing method for intelligent customer service, characterized in that: include: S11: parsing a sentence input by a user, wherein the parsing includes deep semantic understanding; S12: Determine the query target and query conditions according to the parsing result; S13: According to the query target and query conditions, a function call chain is constructed, and the state flow is driven by combining the business processing metamodel, and the corresponding business function is called and executed; wherein the business processing metamodel includes all business states and the transfer logic between business states.
2. The multi-task wizard dialogue processing method for intelligent customer service according to claim 1, characterized in that: Step S11 specifically includes: S111: classifying potential query targets of users in a preset service scenario, and setting corresponding basic semantic units for each type of query target; S112: Use the pre-trained neural network model to fine-tune the results of S111 to obtain the key semantic slot content; S113: Systematically sort out the knowledge points of the semantic slot content, and add keywords and subject terms according to the preset knowledge base; S114: Create a multi-dimensional and multi-level index structure for the same knowledge point; S115: Convert the preset knowledge base into a structured knowledge graph and build a triple model with entities, attributes and relationships as the core; S116: According to the index structure, the triple model is used to retrieve the semantic slot content of step 12 to obtain corresponding parsed content.
3. The multi-task wizard dialogue processing method for intelligent customer service according to claim 1, characterized in that: The parsed result includes one or more elements of subject, predicate, object, attributive, adverbial, and complement; step S12 specifically includes: S121: extracting noun phrases from the parsed results; S122: extracting predicates from the parsed results, and identifying events based on the extracted predicates and noun phrases; S123: Analyze the parallel and limiting relationships between noun phrases, perform cascade processing on limiting relationships, and perform cross multiplication operations on parallel relationships; construct a semantic expression containing one or more of the subject, predicate, object, attributive, adverbial, and complement; S124: construct a grammatical component tree according to the results of steps S121-S123; perform a search according to the grammatical component tree to generate a query target and a query condition.
4. The multi-task wizard dialogue processing method for intelligent customer service according to claim 3, characterized in that: Step S124 specifically includes: S1241: construct a grammatical component tree according to the results of steps S121-S123; establish the subject and object as the query target through minimum recursive traversal; S1242: Perform cross product processing on related adverbials and complements to generate query conditions.
5. The multi-task wizard dialogue processing method for intelligent customer service according to claim 1, characterized in that: Step S13 specifically includes: S131: searching for a query target function from a preset function library according to the determined query target. If the query target function is not found, triggering a preset fallback function; if the query target function is found, executing the following steps; S132: Decomposing the query target and the query condition and mapping them to a function call chain; S133: Using the business processing metamodel to drive the state flow, calling and executing the corresponding business function according to the function call chain.
6. The multi-task wizard dialogue processing method for intelligent customer service according to claim 5, characterized in that: The S133 specifically includes: The context information of the conversation is embedded in the function call chain, and the preset business processing metamodel is used to perform node judgment; the corresponding business call process is triggered according to the processing node where the conversation is located.
7. The multi-task wizard dialogue processing method for intelligent customer service according to claim 6, characterized in that: The business processing metamodel includes several key business processing nodes. When the query target or query condition includes multiple tasks, S133 specifically includes: The decomposed query conditions and query parameters are input into the processing nodes of each business processing metamodel, triggering the calling of the corresponding business function, and executing the called business function.
8. A multi-task wizard-style dialogue processing system for intelligent customer service, characterized in that: include: A syntax analysis module, used to parse the sentences input by the user, wherein the parsing includes deep semantic understanding; The target and condition identification module is used to determine the query target and query condition according to the parsing result; A dynamic task mobilization module is used to build a function call chain according to the query target and query conditions, and call and execute the corresponding business function according to the function call chain; The metamodel-driven state flow management module is used to control the flow of business states using the business processing metamodel; the business processing metamodel includes all business states and the transfer logic between business states.
9. The multi-task wizard dialogue processing system of intelligent customer service according to claim 8, characterized in that: The dynamic task mobilization module is specifically used for: According to the determined query target, the query target function is retrieved from the preset function library. If the query target function is not retrieved, the preset fallback function is triggered; if the query target function is retrieved, the query target and the query condition are disassembled and mapped to the function call chain; according to the function call chain, the corresponding business function is called and executed.
10. The multi-task wizard dialogue processing method for intelligent customer service according to claim 9, characterized in that: The business processing metamodel includes several key processing nodes. The dynamic task mobilization module embeds the context information of the conversation into the function call chain. The metamodel-driven state flow management module determines the processing node where the conversation is located. The dynamic task mobilization module triggers the corresponding business call process according to the processing node where the conversation is located. When processing multiple tasks, the corresponding business function is triggered and executed according to each task processing node.
Citation Information
Cited By
Online and offline business collaborative optimization method and system of service platform
CN120430759A
Online and offline business cooperation optimization method and system of service platform
CN120430759B
