An intelligent logistics customer service question and answer method, device, equipment and medium

Abstract

This invention relates to the field of intelligent interaction and logistics management technology, and discloses an intelligent logistics customer service question-and-answer method, apparatus, equipment, and medium, comprising: receiving a logistics inquiry request; identifying the user's inquiry intent and set of business elements in the logistics inquiry request; retrieving logistics nodes, business rules, and historical cases that match the user's inquiry intent and set of business elements; constructing a response framework for the logistics inquiry request; retrieving real-time logistics status data and generating candidate answers using the response framework; extracting intent type features from the user's inquiry intent and filtering priority determination elements from the set of business elements; determining the logistics business priority based on the priority determination elements; calculating a weighted score for each candidate answer based on the intent type features and the logistics business priority; and selecting the candidate answer with the highest weighted score as the target customer service answer. This invention can improve the accuracy of intelligent logistics customer service question-and-answer.

Description

Technical Field

[0001] This invention relates to the field of intelligent interaction and logistics management technology, and in particular to an intelligent logistics customer service question-and-answer method, device, equipment and medium. Background Technology

[0002] The intelligent logistics industry is developing rapidly, and users are constantly demanding more real-time, accurate, and scenario-adaptable logistics consultations. Intelligent logistics customer service Q&A systems have become a core platform for logistics companies to improve efficiency and reduce costs. These systems can be applied to various terminals such as logistics apps and mini-programs, adapting to the consultation needs of multiple logistics scenarios including road transport, express delivery, and warehousing and distribution. They rely on natural language processing technology to analyze and respond to user inquiries. However, the complex business elements and dynamically changing states of the logistics industry place higher demands on the professional adaptability of this system. Therefore, to meet the logistics industry's needs for efficient and intelligent customer service, it is necessary to customize intelligent logistics customer service Q&A methods to improve their accuracy.

[0003] Existing technologies only perform shallow keyword matching for logistics consultation requests, without semantic deconstruction and integration of logistics business data, making it impossible to accurately extract consultation intent and business elements. They only perform global fuzzy searches in knowledge graphs, failing to achieve precise matching of logistics nodes, business rules, historical cases, scenarios, business processes, and characteristics, resulting in untargeted search results. They do not build a dedicated response logic framework for consultation requests, simply piecing together retrieved knowledge content. They do not retrieve real-time logistics status data, relying solely on static knowledge to generate answers, failing to produce differentiated candidate responses. Furthermore, they only use semantic matching as a single metric for scoring, without combining intent type characteristics and business priorities to calculate scores, and without selecting the optimal answer, leading to low accuracy in intelligent logistics customer service Q&A. Summary of the Invention

[0004] This invention provides a method, apparatus, equipment, and medium for intelligent logistics customer service Q&A, in order to solve the problem of low accuracy in intelligent logistics customer service Q&A.

[0005] Firstly, an intelligent logistics customer service Q&A method is provided, including: Receive logistics consultation requests initiated by target users, perform semantic deconstruction and business element mapping on the logistics consultation requests, and obtain a set of user consultation intent and logistics business elements. Based on a preset hierarchical similarity threshold filtering strategy, the system retrieves target logistics nodes, business matching rules, and feature-related historical cases from a preset logistics knowledge graph that match the user's consultation intent and the set of logistics business elements, respectively, based on node matching, business matching, and feature matching. Based on the target logistics node, the business adaptation rules, and the historical cases associated with the features, a perception and response framework for the logistics consultation request is constructed in conjunction with the preset logistics scenario self-adaptation logic. The system filters out identifying business elements with target retrieval attributes from the set of logistics business elements, retrieves real-time logistics status data corresponding to the logistics inquiry request from the preset real-time logistics business database based on the identifying business elements, integrates the real-time logistics status data into the perception and response framework, and uses the perception and response framework to generate multiple differentiated candidate answers. Extract the intent type features from the user's inquiry intent, and filter the priority determination elements in the logistics business element set. Determine the logistics business priority of the logistics inquiry request based on the priority determination elements. Calculate the weighted score of each differentiated candidate answer based on the intent type features and the logistics business priority. The candidate answer with the highest weighted score is selected as the target customer service answer.

[0006] Secondly, an intelligent logistics customer service Q&A device is provided, including: The consultation intent and element extraction module is used to receive logistics consultation requests initiated by target users, perform semantic deconstruction and business element mapping processing on the logistics consultation requests, and obtain a set of user consultation intent and logistics business elements. The logistics knowledge graph retrieval module is used to retrieve, according to a preset hierarchical similarity threshold filtering strategy, target logistics nodes that match the user's consultation intent and the set of logistics business elements respectively in terms of node matching, business matching, and feature matching, as well as historical cases of business adaptation rules and feature association from the preset logistics knowledge graph. The perception and response framework construction module is used to construct the perception and response framework for the logistics consultation request based on the target logistics node, the business adaptation rules and the historical cases associated with the features, combined with the preset logistics scenario self-adaptation logic. The candidate answer generation module is used to filter out the identifying business elements with target retrieval attributes in the set of logistics business elements, retrieve the real-time logistics status data corresponding to the logistics consultation request from the preset real-time logistics business database according to the identifying business elements, integrate the real-time logistics status data into the perception response framework, and use the perception response framework to generate multiple differentiated candidate answers. The candidate answer weighted score calculation module is used to extract the intent type features from the user's consultation intent, filter the priority determination elements in the logistics business element set, mark the logistics business priority of the logistics consultation request according to the priority determination elements, and calculate the weighted score of each differentiated candidate answer according to the intent type features and the logistics business priority. The target customer service response filtering module is used to filter the differentiated candidate responses with the highest scores among the weighted calculated scores as the target customer service responses.

[0007] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described intelligent logistics customer service Q&A method.

[0008] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the aforementioned intelligent logistics customer service Q&A method.

[0009] The solutions implemented by the aforementioned intelligent logistics customer service question-and-answer methods, devices, equipment, and media can achieve customized processing of the entire process from parsing the consultation request to outputting the target answer through the client. This effectively improves the accuracy of intent parsing and knowledge retrieval, makes the answer both real-time and logically coherent, and can scientifically quantify and evaluate candidate answers to select the optimal answer that fits the core needs. This can solve the problem of low accuracy in intelligent logistics customer service question-and-answer. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of an application environment for an intelligent logistics customer service question-and-answer method according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating an intelligent logistics customer service question-and-answer method according to an embodiment of the present invention; Figure 3 yes Figure 2 A flowchart illustrating a specific implementation method of step S1; Figure 4 yes Figure 2 A flowchart illustrating a specific implementation method of step S3; Figure 5 This is a schematic diagram of the structure of an intelligent logistics customer service Q&A device in one embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention; Figure 7 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0012] 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, not all, of the embodiments of the present invention. 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.

[0013] The intelligent logistics customer service question-and-answer method provided in this embodiment of the invention can be applied to, for example... Figure 1 In this application environment, the client communicates with the server via a network. The server can utilize the client to achieve customized processing throughout the entire process, from parsing the inquiry request to outputting the target answer. This effectively improves the accuracy of intent analysis and knowledge retrieval, ensuring the answer is both real-time and logically coherent. It also allows for scientific quantification and evaluation of candidate answers, selecting the optimal answer that best meets the core needs, thus addressing the issue of low accuracy in intelligent logistics customer service Q&A. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a dedicated server or a server cluster consisting of multiple servers. The invention will be described in detail below through specific embodiments.

[0014] Please see Figure 2 As shown, Figure 2 A flowchart illustrating the intelligent logistics customer service question-and-answer method provided in this embodiment of the invention includes the following steps: S1. Receive a logistics consultation request initiated by the target user, perform semantic deconstruction and business element mapping processing on the logistics consultation request, and obtain the user's consultation intent and a set of logistics business elements.

[0015] In this embodiment of the invention, the target user can be an individual or enterprise that has logistics-related consultation needs in logistics scenarios such as highways, express delivery, and warehousing and distribution. Such entities will initiate relevant consultation behaviors through intelligent logistics customer service interaction terminals such as logistics APPs, mini-programs, web pages, and self-service inquiry machines. The logistics consultation request can be various logistics-related consultation information sent by the entity with logistics consultation needs through the above-mentioned intelligent logistics customer service interaction terminals. It is mostly presented in text form and can cover various logistics business-related requests such as waybill tracking, delivery time inquiry, lost package compensation consultation, address modification application, and freight standard consultation.

[0016] In detail, an intelligent logistics customer service Q&A system is a core technology system developed based on three core technology areas: intelligent logistics, natural language processing, and intelligent customer service. It is customized for the characteristics of the logistics industry, including highways, express delivery, and warehousing and distribution. It serves as a core technology carrier for logistics companies to improve customer service efficiency and reduce labor costs. The system uses intelligent logistics customer service interactive terminals such as logistics apps, mini-programs, web pages, and self-service kiosks as the interactive entry points for user inquiries. It can achieve intelligent processing of the entire process of user-initiated logistics inquiries, covering core aspects such as semantic deconstruction mapping, multi-level knowledge graph retrieval, real-time business data fusion, dynamic weighted evaluation, and accurate natural language output. Ultimately, it achieves accurate parsing, real-time response, and scenario-based output of user logistics inquiries, improving the efficiency, accuracy, real-time nature, and scenario adaptability of logistics customer service Q&A.

[0017] Specifically, the intelligent logistics customer service Q&A system relies on intelligent logistics customer service interactive terminals such as logistics apps, mini-programs, web pages, and self-service inquiry machines in logistics scenarios such as highways, express delivery, and warehousing and distribution to build a dedicated information receiving channel. It captures the operation behavior of target users initiating logistics consultation requests on such interactive terminals, fully captures the relevant information of the logistics consultation requests submitted by target users on the terminals, and directly inputs this information into the processing module of the intelligent logistics customer service Q&A system.

[0018] In this embodiment of the invention, the user's consultation intent can be various core logistics consultation requests obtained after the logistics consultation request is classified by a pre-trained logistics intent classification model. These requests can cover specific consultation purposes in logistics scenarios such as waybill tracking and timeliness inquiry, lost package compensation consultation, timeliness complaint, address modification application, and freight consultation. This is the core demand of the user initiating the logistics consultation request. The set of logistics business elements can be a set formed by structurally integrating the core business entities extracted after the logistics consultation request has been segmented, part-of-speech tagging, stop word removal, and logistics-specific named entity recognition. This set can include various key logistics business entity information related to the logistics consultation request, such as waybill number, waybill type, delivery timeliness, sender and receiver address, current logistics node, core logistics node, and dedicated delivery network.

[0019] In this embodiment of the invention, reference is made to Figure 3 As shown, the semantic deconstruction and business element mapping processing of the logistics consultation request yields a set of user consultation intent and logistics business elements, including: S31. Perform lexical and semantic analysis on the logistics consultation request to obtain a logistics consultation word segmentation feature set; S32. Extract business entities from the logistics consultation word segmentation feature set to obtain a logistics business entity set; S33. Call the logistics consultation intent feature library in the pre-trained logistics intent classification model to perform feature matching analysis on the logistics business entity set, and determine the intent category of the feature results obtained after matching to obtain the user consultation intent. S34. Extract the intent feature dimension of the user's consultation intent and the entity feature dimension of the logistics business entity set, align the intent feature dimension and the entity feature dimension, and perform structured association and integration on the aligned intent entity feature pairs to obtain the element mapping relationship between the user's consultation intent and the logistics business entity set. S35. Based on the element mapping relationship, perform element fusion processing on the user consultation intent and the logistics business entity set to obtain a logistics business element set.

[0020] In detail, the logistics consultation word segmentation feature set can be a collection of effective core words and corresponding part-of-speech features formed after word segmentation, part-of-speech tagging, and stop word removal operations are performed on the logistics consultation request. This collection can extract key word information in the logistics consultation request and provide basic word feature support for subsequent processing such as entity recognition and intent classification. The intelligent logistics customer service question and answer system performs word segmentation operation to split the text content of the received logistics consultation request, performs part-of-speech tagging operation to tag the corresponding part of speech for each word after splitting, and then performs filtering operation to remove stop words. The processed effective word segmentation and corresponding part-of-speech features are integrated to form the logistics consultation word segmentation feature set.

[0021] Specifically, the logistics business entity set can be a collection formed by integrating various core logistics business-related entity information after extracting business entities from the logistics consultation word segmentation feature set. This set can cover key logistics business entity information related to logistics consultation requests, such as waybill number, sender and receiver address, logistics node, and delivery time. The intelligent logistics customer service Q&A system calls the logistics-specific named entity recognition model to perform extraction operations on the logistics consultation word segmentation feature set, and filters and extracts core logistics business entities such as waybill number, sender and receiver address, logistics node, and delivery time from the logistics consultation word segmentation feature set. The extracted core logistics business entities are then integrated to form the logistics business entity set.

[0022] Furthermore, the pre-trained logistics intent classification model can be a natural language processing model trained and optimized based on logistics industry-specific consultation scenarios. It incorporates intent recognition algorithms and feature matching logic specific to the logistics field, making it ideal for classifying and determining various intents in logistics consultation scenarios. The logistics consultation intent feature library can be a set of feature data built into the pre-trained logistics intent classification model, including feature dimensions and feature identifiers corresponding to specific logistics consultation intents such as tracking queries, lost item claims, timeliness complaints, address modifications, and freight inquiries. This provides a reference for feature matching analysis of logistics business entity sets. The feature results can be used to compare the logistics business entity sets with the... The feature matching data generated by the logistics consultation intent feature library after feature matching includes matching-related information such as the feature fit and feature association dimension between the two. The logistics consultation intent feature library built into the pre-trained logistics intent classification model is called to perform feature matching analysis on the various business entity features of the logistics business entity set and the various types of intent features in the logistics consultation intent feature library. The feature results obtained after this matching are used to determine the intent category according to the judgment rules of the pre-trained logistics intent classification model. The category information that matches the feature results is extracted from various logistics consultation intent categories to obtain the consultation intent of the corresponding user's logistics consultation request.

[0023] Furthermore, the intent feature dimension can be various feature attributes and analysis dimensions corresponding to the user's consultation intent, covering feature consideration dimensions related to logistics consultation intent such as type identification, scenario orientation, and core needs; the entity feature dimension can be the feature attributes and analysis dimensions corresponding to each business entity within the logistics business entity set, including the identification attributes of business entities such as waybill number, delivery and receipt address, and logistics node, as well as feature consideration dimensions related to logistics business entities such as business association and scenario adaptation; the intent entity feature pair can be a one-to-one combination of feature dimensions formed after the intent feature dimension and entity feature dimension are aligned, including mutually matching intent feature dimension items and entity feature dimension items; element mapping The relationship can be a feature-dimensional association between user consultation intent and logistics business entity set, reflecting the association logic and matching basis between consultation intent and each business entity. The process involves extracting the corresponding intent feature dimension from the user consultation intent and the corresponding entity feature dimension from the logistics business entity set. The extracted intent feature dimension and entity feature dimension are then aligned to clarify the matching relationship between the two types of feature dimensions. The resulting intent-entity feature pairs are then structurally integrated, clarifying the association logic of each intent-entity feature pair and integrating relevant feature dimension information according to structured rules to obtain the element mapping relationship between user consultation intent and logistics business entity set.

[0024] Furthermore, based on the element mapping relationship, the consultation elements contained in the user consultation intent and the business elements contained in the logistics business entity set are decomposed. Referring to the corresponding association logic established in the element mapping relationship, the user consultation elements and logistics business elements are matched, superimposed, and supplemented according to the association relationship. The information items that are related to each other in the two types of elements are integrated, duplicate elements are integrated, missing related items are filled in, the structure and format of the elements are standardized, the expression norms and data levels of the elements are unified, and the element fusion processing is completed to generate a set of logistics business elements that contains all related elements of the user consultation intent and the logistics business entity set.

[0025] S2. Based on the preset hierarchical similarity threshold filtering strategy, retrieve target logistics nodes, business matching rules and feature association historical cases from the preset logistics knowledge graph that match the user's consultation intent and the set of logistics business elements respectively in terms of node matching, business matching and feature matching.

[0026] In this embodiment of the invention, the preset hierarchical similarity threshold can be a quantitative value pre-configured for each level of the multi-level logistics knowledge graph to determine the matching fit. Different threshold values ​​can be configured for the logistics basic rules layer, logistics business process layer, and logistics historical case layer to define the qualification standards for matching content within each level. The preset logistics knowledge graph can be a multi-level knowledge system customized for the logistics industry, typically divided into three layers: the logistics basic rules layer, the logistics business process layer, and the logistics historical case layer. It covers general logistics rules, basic processes, process-related nodes for subdivided logistics scenarios, and specific business specifications. This includes historical logistics consultation cases, solutions, and other logistics-related knowledge. The scenario matching involves associating the user's consultation intent and logistics business elements with relevant content in the logistics knowledge graph, tailored to the specific logistics scenario of the user's consultation. This matching process will align with specific logistics consultation scenario characteristics such as fresh produce express delivery, lost package claims, timeliness complaints, and address modifications. The business matching involves accurately matching the logistics business elements extracted by the user with the core content of the logistics business process layer in the logistics knowledge graph, based on the professional business logic of the logistics industry. The matching dimensions will cover the entire logistics business process. The system includes aspects related to actual logistics operations, such as process-related nodes and specific business operation specifications. Feature matching involves matching the type characteristics of user inquiries, the identifying characteristics of logistics business elements, and the case characteristics of historical cases in the logistics knowledge graph to identify historical processing cases that share common characteristics with the current logistics inquiry. The target logistics node is a key node in the logistics business process with actual business operation attributes, typically including core nodes such as sorting centers, distribution outlets, and transportation link transfer points. It represents a key location in the logistics process that matches the user's inquiries and logistics business elements. Business adaptation rules are specific business operation specifications and processing guidelines developed by the logistics industry for different logistics scenarios and waybill types. These typically cover priority delivery rules for fresh and urgent shipments, execution rules for different delivery times, and rules for handling various logistics problems, adapting to different logistics business needs. Historical cases related to features can be actual cases of logistics inquiries and problem-solving handled in the past by the logistics industry. These are typically historical logistics inquiry processing cases and logistics problem solution cases that share similar characteristics with the current user's inquiries and logistics business elements, providing practical supplementary reference for responding to the current logistics inquiry.

[0027] In this embodiment of the invention, the step of retrieving target logistics nodes, business matching rules, and feature association historical cases from a preset logistics knowledge graph based on a preset hierarchical similarity threshold filtering strategy includes: The user's consultation intent is decomposed into contextual features to obtain an intent scenario feature set. The logistics business element set is decomposed into business features to obtain a business attribute feature set. The intent scenario feature set and the business attribute feature set are fused to obtain a target retrieval feature set. Using the retrieval dimensions corresponding to the three-level structure in the preset logistics knowledge graph, hierarchical feature matching is performed on the target retrieval feature set to obtain the matching similarity value of each level. The matching similarity values ​​of each level are compared with the preset level similarity threshold to filter out candidate content whose matching similarity values ​​are greater than or equal to the corresponding level similarity threshold. The candidate content is mapped to process-related nodes to obtain a node-related candidate set; Identify the matching relationship between the node association candidate set and the logistics business element set, and select target logistics nodes, business adaptation rules and feature association historical cases from the node association candidate set according to the matching relationship.

[0028] In detail, the intent scenario feature set can be a feature set tailored to specific logistics consultation scenarios, formed by integrating features after performing scenario-based feature decomposition on the user's consultation intent. It may include feature dimension information corresponding to logistics consultation scenarios such as tracking, lost package claims, timeliness complaints, and address modifications. The business attribute feature set can be a feature set matching core logistics business elements, formed by integrating features after performing business-based feature decomposition on the set of logistics business elements. It may include feature dimension information corresponding to logistics business entities such as waybill number, sender / receiver address, logistics node, delivery timeliness, and waybill type. The target retrieval feature set can be an integrated retrieval feature set combining logistics consultation scenario features and logistics business element features, formed by performing feature fusion on the intent scenario feature set and the business attribute feature set. It can simultaneously carry the scenario attributes of the user's core consultation intent and key logistics business. The elements include attribute information; scenario-based feature decomposition can be based on the logistics consultation sub-scenarios corresponding to the user's consultation intent, decomposing all feature dimensions that match the logistics consultation scenario under the intent, sorting and integrating the decomposed scenario feature dimensions to form an intent scenario feature set; business-based feature decomposition can be based on the various core business entities in the logistics business element set, decomposing all feature dimensions that match the core logistics business under each business entity, sorting and integrating the decomposed business feature dimensions to form a business attribute feature set; feature fusion can align all feature dimensions in the intent scenario feature set with all feature dimensions in the business attribute feature set, carry out feature dimension association integration and feature information complementarity fusion on the aligned feature dimensions, sort and integrate all fused feature dimensions and feature information to form a target retrieval feature set.

[0029] Specifically, the retrieval dimensions corresponding to the three-tier structure can be: logistics general rules and basic process-related retrieval dimensions corresponding to the logistics basic rules layer; detailed logistics scenario process association nodes and dedicated business adaptation rules-related retrieval dimensions corresponding to the logistics business process layer; and historical logistics consulting and processing cases and solution-related retrieval dimensions corresponding to the logistics historical case layer. Each of these three retrieval dimensions corresponds to a three-tier structure of the logistics knowledge graph and has its own exclusive retrieval matching direction. The matching similarity value at each level can be: the logistics general rules and basic process-related matching values ​​obtained after feature matching of the target retrieval feature set in the logistics basic rules layer; the process association nodes and business adaptation rules-related matching values ​​obtained after feature matching in the logistics business process layer; and the feature-related historical cases and past solution-related matching values ​​obtained after feature matching in the logistics historical case layer. Each level of the logistics knowledge graph... After feature matching is completed at each level, the relevant knowledge content can become the matching similarity value for that level. Based on the exclusive retrieval dimensions corresponding to the three-level structure of logistics basic rules, logistics business process, and logistics historical case in the pre-built logistics knowledge graph, these dimensions serve as the core basis and specific direction for feature matching of the target retrieval feature set. Layered feature matching projects the target retrieval feature set onto the respective retrieval dimensions of the three-level structure of the logistics knowledge graph. Targeted feature matching operations are performed according to the knowledge attributes and feature requirements of each retrieval dimension. The matching operation will match the exclusive attributes of each level's retrieval dimension with the feature dimensions of the target retrieval feature set. The result is that after completing the feature matching operation of the target retrieval feature set within each level's retrieval dimension of the logistics knowledge graph, the knowledge content that matches the target retrieval feature set in each level is extracted, forming the matching similarity value corresponding to each level.

[0030] Furthermore, the candidate content can be various logistics-related knowledge content obtained by matching the search feature vector with the features of each level of the multi-level logistics knowledge graph. This includes the content obtained by matching the basic logistics rules, business process nodes, and feature-related historical cases of each level. When performing the comparison operation, the matching similarity value calculated by each level of the multi-level logistics knowledge graph needs to be compared with the pre-configured level similarity threshold of the corresponding level. When performing the filtering operation, logistics knowledge content with matching similarity values ​​greater than or equal to the preset level similarity threshold of the corresponding level needs to be retained in the matching results of each level. This completes the filtering operation of candidate content at each level.

[0031] Furthermore, the node association candidate set can be a collection of logistics process association node classes related to logistics consultation requests, formed by integrating the candidate content at each level of the logistics knowledge graph after completing the process association node mapping operation. It can cover various matched logistics process association nodes and related business information, business adaptation rules, feature association historical case points, and other related information. The mapping operation involves matching the candidate content selected at each level of the logistics knowledge graph to the corresponding logistics business process association nodes one by one according to the business process logic of the logistics industry and the association attributes of the process association nodes. The information related to each process association node in the candidate content is sorted and collected into the corresponding process association node. Then, all the logistics process association nodes after the mapping and the collected information corresponding to each node are sorted and integrated in a structured manner to obtain the node association candidate set.

[0032] Furthermore, the matching relationship can be the fitting association attributes formed between various contents in the node association candidate set and core elements in the logistics business element set, such as waybill number, waybill type, delivery time, delivery and receipt address, and logistics node. It typically includes attributes that reflect the relationship between the two, such as similarity at the feature level, correspondence and adaptability at the business level, and matching fit at the scenario level. This is the core reference for judging whether the content in the node association candidate set matches the user's actual logistics consultation needs. Identification involves comparing each core business element in the logistics business element set with the fitting association attributes between each item in the node association candidate set and the logistics business element set, and then... The specific correlation attributes between the two, such as feature similarity, business adaptability, and scenario fit, are clarified. The core screening criterion is to use the various matching correlation attributes between the identified node correlation candidate set and the logistics business element set as the core screening criteria. Based on these correlation attributes, the screening criteria for relevant content in the node correlation candidate set are defined. The screening process is to select the logistics node type content, business adaptability type content, and feature correlation historical case type content with the highest degree of fit with the logistics business element set, the strongest business adaptability, and the best scenario fit from the node correlation candidate set according to the defined screening criteria. These are respectively determined as the target logistics node, business adaptability rule, and feature correlation historical case.

[0033] S3. Based on the target logistics node, the business adaptation rules, and the historical cases associated with the features, construct a perception and response framework for the logistics consultation request in conjunction with the preset logistics scenario self-adaptation logic.

[0034] In this embodiment of the invention, the preset logistics scenario self-adaptation logic can be a set of rules and judgment systems pre-configured in the system for the construction of the perception and response framework, specifically adapted to various sub-scenarios of logistics. It typically includes logistics scenario classification mapping rules, activation and hiding rules for dynamically extended nodes, and adaptation priority rules for framework logic. It may also include execution logic that automatically matches corresponding logistics nodes, business adaptation rules, and historical case supplementation dimensions based on the actual scenario characteristics of logistics consultation. This logic system will pre-set exclusive mapping relationships between each scenario and logistics node type, business adaptation rules, and case supplementation dimensions around different sub-scenarios of logistics such as road transport, express delivery, fresh produce express shipments, and lost shipment claims. It will also pre-set activation conditions for various scalable logistics nodes and the core orientation of the framework's main logic in different scenarios. Furthermore, based on scenario characteristics such as waybill type and consultation intent type extracted from the logistics consultation request, it can automatically call corresponding rules to adjust the node composition, business rule constraint strength, and historical case supplementation weight of the perception and response framework. The logic system can be flexibly expanded and its rules adjusted according to the business updates and new scenario requirements of logistics enterprises, thereby ensuring that the perception and response framework built for each logistics consultation request can accurately fit the specific logistics scenario requirements. The perception and response framework can be a dedicated structured response logic framework built for a single logistics consultation request. It is constructed by combining process-related nodes, business adaptation rules, and feature-related historical cases that match the logistics consultation request. The process-related nodes serve as the main logic of the framework, the business adaptation rules serve as the logical constraints of the framework, and the feature-related historical cases serve as supplementary references to the framework, realizing the dynamic association and combination of three types of logistics business knowledge. At the same time, each logical node of the framework reserves a real-time data injection interface, which can provide a foundation for the injection of real-time status data of subsequent logistics business. Based on this framework, the dynamic association and combination of process-related nodes, business adaptation rules, feature-related historical cases, and real-time business data can be realized, so that the generated logistics customer service answers have logic and contextual coherence that fit the logistics consultation scenario.

[0035] In this embodiment of the invention, reference is made to Figure 4 As shown, the step of constructing a perception and response framework for the logistics consultation request based on the target logistics node, the business adaptation rules, and the historical cases associated with the features, combined with a preset logistics scenario self-adaptation logic, includes: S41. The target logistics node is sorted out to obtain the logistics node link; S42. Logically constrain the logistics node link according to the business adaptation rules to obtain the logistics node logical link, and configure logical nodes for the logistics node logical link according to the preset logistics scenario self-adaptation logic. S43. Decompose the historical cases associated with the features into logistics scenario dimensions to obtain supplementary case-based evidence; S44. Integrate the case-based supplementary criteria into each logical node of the logistics node logical link to obtain the perception response framework.

[0036] In detail, a logistics node link can be a sequence of core logistics nodes that are sequentially connected in the logistics and distribution process, formed around a target logistics node. It includes the flow order of each logistics node and the upstream and downstream relationships during the distribution process. The sorting is directly aimed at the target logistics node, relying on the distribution process rules of the logistics business, to accurately locate the upstream and downstream related distribution link nodes in the distribution link to which the target logistics node belongs, clarify the distribution flow logic and connection relationship between each related distribution link node and the target logistics node, and organize and arrange these related distribution link nodes in an orderly manner according to the actual delivery sequence.

[0037] Specifically, a logistics node logical link can be a sequence of logistics nodes that combines the flow relationship of delivery nodes with the requirements of business rules after being logically constrained by business adaptation rules. It integrates the business adaptation rules of logistics sub-scenarios, clarifies the flow conditions, operation standards and connection specifications of each node in the logistics node link, and ensures that the link relationship between nodes meets the specific adaptation requirements of logistics business. The constraints are based on the self-adaptive logic of the logistics scenario corresponding to the logistics sub-scenarios. It limits and standardizes the delivery nodes and the flow relationship between nodes in the logistics node link at the logical level. According to the self-adaptive logic of the logistics scenario, it clarifies the operation requirements, flow premises and connection criteria of each node in the logistics node link. It makes compliance adjustments to the node arrangement or flow relationship in the logistics node link that does not conform to the self-adaptive logic of the logistics scenario. It standardizes the overall flow logic of the logistics node link according to the self-adaptive logic of the logistics scenario, so that the node sequence and the connection relationship between nodes in the logistics node link fully meet the requirements of the self-adaptive logic of the logistics scenario.

[0038] Furthermore, supplementary case studies can be scenario-based case information derived from breaking down historical cases related to features according to logistics scenario dimensions. This information is adapted to the current logistics consulting scenario and can serve as supplementary reference for responses. It includes the processing ideas, practical methods, and key points of historical consultations under each logistics scenario dimension. The breakdown involves decomposing and analyzing the overall content of historical cases related to features at the logistics scenario level, combined with the actual logistics scenario dimension corresponding to the logistics consulting. This process uncovers the core processing information, problem-solving methods, and practical execution points corresponding to historical cases related to features under each logistics scenario dimension. The case information of each logistics scenario dimension after decomposition and analysis is then screened and integrated to extract case content that matches the current logistics consulting scenario. The integrated scenario-based case information is then organized into a standardized content format that can be directly used as a reference for responses.

[0039] Furthermore, logical nodes can be the core logistics nodes that constitute the logistics distribution flow logic in the logistics node logical link. They are key nodes in the link with independent business operation attributes and flow connection functions, bearing the constraints of business adaptation rules under the corresponding logistics scenario and the flow logic between nodes. Integration involves matching historical case information corresponding to the logistics scenario dimension in the case-based supplementary basis for each logical node in the logistics node logical link, mining the processing ideas and key points related to the business operation and flow connection of each logical node in the case-based supplementary basis, and associating these adapted case-based supplementary basis contents with each logical node in the logistics node logical link. The case-based supplementary basis is embedded as supplementary reference content for each logical node, so that each logical node has both the flow logic constrained by business adaptation rules and the corresponding scenario-based case reference basis. Through this integration operation, the core flow logic of the logistics node logical link and the scenario-based reference information of the case-based supplementary basis are integrated to form a perception and response framework that has both logical coherence and scenario reference.

[0040] S4. Filter the identifying business elements with target retrieval attributes in the set of logistics business elements, retrieve the real-time logistics status data corresponding to the logistics consultation request from the preset real-time logistics business database according to the identifying business elements, integrate the real-time logistics status data into the perception response framework, and use the perception response framework to generate multiple differentiated candidate answers.

[0041] In this embodiment of the invention, the target retrieval attribute can be an attribute feature that has unique pointing and precise matching, and can be used as a retrieval basis to retrieve the corresponding real-time status data of logistics business from the real-time logistics business database. It is an exclusive attribute feature adapted to the retrieval of real-time logistics business data. The identifying business element is usually the core business element with such target retrieval attribute in the set of logistics business elements. It can include unique waybill numbers, core logistics nodes, dedicated delivery outlets, and other logistics business elements with unique identification that can be directly used as retrieval keywords. Based on its own target retrieval attribute, such elements can achieve accurate and efficient retrieval of the corresponding real-time status data in the real-time logistics business database.

[0042] In this embodiment of the invention, the step of filtering the identifying business elements with target retrieval attributes from the set of logistics business elements includes: Based on the preset retrieval attribute dimensions, the feature information of each business element in the set of logistics business elements is classified and decomposed to obtain the retrieval attribute feature set corresponding to each business element; Extract the dimensional feature requirements corresponding to the target retrieval attributes, and quantify and decompose the dimensional feature requirements dimension by dimension according to the retrieval attribute dimensions to obtain the quantitative evaluation index corresponding to each retrieval attribute dimension, and assign corresponding index weights to each quantitative evaluation index. Using the quantitative evaluation indicators and the corresponding indicator weights, the dimensional feature matching calculation is performed on the retrieval attribute feature set corresponding to each business element to obtain the matching score of each retrieval attribute dimension. The matching scores of each search attribute dimension are weighted and summed according to the corresponding indicator weights to obtain the dimension fit value of each business element. All business elements are sorted in descending order of their dimensional fit value to obtain an adaptation ranking result. The business elements in the preset first-place interval of the adaptation ranking result are identified as identifiable business elements with target retrieval attributes.

[0043] In detail, the preset search attribute dimensions can be various search attribute division dimensions set in logistics operations for retrieving real-time data, typically including attribute division dimensions related to real-time logistics database retrieval such as identification dimensions, logistics node dimensions, delivery timeliness dimensions, and waybill type dimensions; business elements can be core logistics business-related information extracted and integrated from user logistics inquiry requests through semantic deconstruction mapping, typically including logistics business-related information such as waybill number, sender and receiver address, logistics node, delivery timeliness, and waybill type; the search attribute feature set can be the corresponding feature information set formed after a single business element is classified and decomposed according to the search attribute dimensions, typically including each business element This involves classifying and decomposing the specific feature information and search adaptation conditions for each business element within the logistics business element set under different search attribute dimensions. Following preset classification standards for various search attribute dimensions, the feature information of each business element is categorized and organized. The various feature information of each business element is then split and divided according to different search attribute dimensions. Finally, the split and divided feature information is integrated and organized to form a search attribute feature set corresponding to each business element. This completes the classification and decomposition of feature information for all business elements in the logistics business element set based on preset search attribute dimensions and the construction of corresponding search attribute feature sets.

[0044] Specifically, dimensional feature requirements can be feature adaptation specifications and matching criteria at each dimension level corresponding to the target retrieval attribute. These typically include feature matching standards, retrieval accuracy requirements, and feature relevance requirements for each dimension, all related to the adaptation of the target retrieval attribute. Quantitative evaluation indicators can be specific evaluation criteria that can be quantified after the dimensional feature requirements are quantitatively decomposed dimension by dimension. These typically include quantifiable retrieval attribute evaluation standards such as numerical judgment standards, proportional matching specifications, and graded evaluation requirements for each retrieval attribute dimension. Indicator weights can be numerical coefficients assigned to each quantitative evaluation indicator to reflect the importance of the indicator. These weights typically include those based on logistics business retrieval needs. The importance of search attribute dimensions is determined by setting decimal coefficients, percentage coefficients, and other weight values ​​corresponding to the importance of the indicators. The dimensional feature requirements corresponding to the target search attribute are extracted. Based on the search attribute dimensions, the dimensional feature requirements are quantitatively decomposed dimension by dimension. The dimensional feature requirements under each search attribute dimension are transformed into quantifiable specific judgment criteria. Quantitative evaluation indicators are formed for each search attribute dimension. Combining the search needs of logistics business with the differences in the importance of search attribute dimensions, the indicator weight allocation operation is performed on each quantitative evaluation indicator, assigning a corresponding weight value to each quantitative evaluation indicator. This completes the dimensional feature requirements are quantitatively decomposed dimension by dimension and the indicator weight allocation of quantitative evaluation indicators.

[0045] Furthermore, the matching score can be a quantitative numerical result obtained after the retrieval attribute feature set of business elements and the corresponding retrieval attribute dimensions have completed dimensional feature matching calculations. It usually includes quantitative data that can intuitively reflect the degree of feature matching, such as the fit value of feature information under each retrieval attribute dimension and the matching conformity score. Using the clearly defined quantitative evaluation indicators and the corresponding indicator weights of each quantitative evaluation indicator, the dimensional feature matching calculation operation is carried out on the retrieval attribute feature set corresponding to each business element. Each feature information in the retrieval attribute feature set is compared and matched with the quantitative evaluation indicators of the corresponding retrieval attribute dimensions one by one. Combined with the indicator weights corresponding to each quantitative evaluation indicator, a weighted operation is performed on the result of each feature comparison and matching verification. The numerical calculation of the weighted result is performed according to the preset calculation rules to obtain a quantitative value that can reflect the degree of matching of each dimension feature, and generate the matching score corresponding to each retrieval attribute dimension. This completes the dimensional feature matching calculation and matching score acquisition of the retrieval attribute feature set based on quantitative evaluation indicators and indicator weights.

[0046] Furthermore, the dimensional fit value can be a quantitative value that comprehensively reflects the overall suitability between business elements and target search attributes, obtained by weighting and summing the matching scores of each search attribute dimension according to the corresponding indicator weights. It typically includes a comprehensive fit value that integrates multi-dimensional matching situations, an overall quantitative score reflecting the suitability of business element search attributes, and other data that can reflect the suitability of business element search. The matching scores of all search attribute dimensions corresponding to each business element and the indicator weights corresponding to the matching scores of each search attribute dimension are collected. The matching score of each search attribute dimension is multiplied by the corresponding indicator weight. The values ​​after weight multiplication of all search attribute dimensions are summed. The comprehensive value after summation is accurately calculated to generate the dimensional fit value corresponding to each business element, thus completing the acquisition of the dimensional fit value based on the weighted summation of the matching scores of search attribute dimensions and indicator weights.

[0047] Furthermore, the adaptive ranking result can be a set of sorted data formed by sorting all business elements in descending order based on their dimension fit values. It typically includes information related to the business elements, their corresponding dimension fit values, and their ranking positions. The preset first-place interval can be a range of ranking positions pre-set based on the actual retrieval needs of logistics business. It typically includes the first few positions after descending sort, the first specific proportion of positions, etc., used to filter highly adaptable business elements. All business elements and their corresponding dimension fit values ​​are extracted. All business elements are sorted in descending order based on their dimension fit values. The sorted business elements, dimension fit values, and ranking position information are then organized and integrated to form the adaptive ranking result. The specific ranking range of the preset first-place interval is defined. Business elements whose ranking falls within the preset first-place interval are selected from the adaptive ranking result. These selected business elements are then defined as identifying business elements with target retrieval attributes. This completes the sorting of business elements based on dimension fit values ​​and the selection and determination of identifying business elements.

[0048] In this embodiment of the invention, the preset real-time logistics business database can be a database system pre-built and deployed by logistics companies to store real-time operational data of logistics business across all dimensions. It can cover various real-time data related to logistics business, such as real-time status of waybills, congestion of transportation links, warehouse inventory information, processing capacity of distribution outlets, and delivery trajectory of couriers. It can achieve accurate data retrieval and retrieval based on identifiable business elements. The real-time logistics status data can be various real-time operational data of logistics business directly corresponding to the specific logistics inquiry request initiated by the user. It can be retrieved from the preset real-time logistics business database using identifiable business elements as search keywords. It can include real-time flow status of the corresponding waybill, real-time traffic conditions of the relevant transportation links, real-time processing capacity of the corresponding distribution outlets, real-time delivery trajectory of the courier responsible for delivery, and other real-time logistics business data content that matches the logistics inquiry request.

[0049] In detail, the system extracts identifiable business elements from the set of logistics business elements, sets these identifiable business elements as search keywords, performs data retrieval operations in the preset real-time logistics business database using these search keywords, matches relevant data corresponding to the user's logistics inquiry request within the preset real-time logistics business database, and retrieves the matched real-time logistics status data from the preset real-time logistics business database.

[0050] Specifically, the real-time data injection interface is preset in the positioning perception and response framework. The real-time logistics status data retrieved from the real-time logistics business database is injected into the corresponding real-time data injection interface of the perception and response framework. Relying on the logical association capability of the perception and response framework, the injected real-time logistics status data is associated with the process association nodes, business adaptation rules, and feature association historical cases within the framework. This allows the real-time logistics status data to form a corresponding match with the logical nodes of the perception and response framework to complete data fusion.

[0051] In this embodiment of the invention, the differentiated candidate answers can be multiple customer service answers generated by performing inference and calculation on the fused data based on the response logic of the perception response framework. They usually differ in terms of expression, information focus, and language conciseness. Each answer is in line with the real-time status of logistics business, has scenario-based and real-time characteristics, and all revolve around the logistics consultation request initiated by the user. The differences lie only in the presentation form and core focus of logistics-related information.

[0052] In this embodiment of the invention, generating multiple differentiated candidate answers using the perceptual response framework includes: The target logistics node of the perception and response framework is used to analyze the logistics business dimension corresponding to the fused data. Based on the logistics business dimension, the fused data is decomposed into node-dimensional data units. The business adaptation rules of the perception and response framework are used to deduce and verify the response information of the nodeized data unit. Based on the deduction and verification results, the response content of the nodeized data unit is logically generated to obtain the node response fragment. By combining the features of the perception and response framework with historical cases, the information focus of the node response fragment is adjusted in a contextualized manner, and the expression style of the adjusted node response fragment is adapted to obtain a contextualized response fragment. Based on the built-in response logic of the perception response framework, the content logic of the scenario-based response fragment is reconstructed to generate multiple differentiated candidate answers.

[0053] In detail, the logistics business dimension can be various logistics business attribute dimensions corresponding to the target logistics nodes within the perception and response framework, covering business analysis dimensions related to logistics node operations such as waybill circulation status, transportation link accessibility, delivery network processing capacity, and courier work progress. The node-based data unit can be the overall data after integrating real-time logistics status data, and the structured business data fragments corresponding to each target logistics node after being decomposed according to the logistics business dimension. It contains specific business data information under each target logistics node that can be directly used to generate response content. Based on the target logistics nodes that have been sorted out in the perception and response framework, the logistics business dimension analysis is carried out on the overall data after integrating real-time logistics status data. The logistics business dimension types that match each target logistics node are sorted out. Then, according to the sorted logistics business dimensions, the node dimension decomposition operation is performed on the integrated overall data. The integrated data is divided into dimensions according to different target logistics nodes. The logistics business dimension related data corresponding to each target logistics node is separated and extracted from the integrated overall data. Then, the extracted single-node data is structured and integrated to form the node-based data unit corresponding to each target logistics node.

[0054] Specifically, the simulation and verification results can be the compliance and rationality judgment information of the response information obtained after performing response information simulation and verification on the nodeized data unit based on the business adaptation rules of the perception and response framework; or they can be the verification feedback information on whether the logistics data in the nodeized data unit matches the logistics business rules and meets the requirements for generating node response information. The node response fragment can be a rule-based and scenario-based response information fragment corresponding to a single logistics target node formed after completing the response information simulation and verification and logical generation of the nodeized data unit; or it can be a basic response content fragment that integrates the actual logistics data of the nodeized data unit and conforms to the business adaptation rules. The simulation and verification results can be obtained by calling the business adaptation rules preset by the perception and response framework and executing the response on the nodeized data unit. The process of deducing the response information involves deriving the direction and content range of the response information that can be transformed from the node-based data unit based on the logistics node flow standards and business operation requirements in the business adaptation rules. Simultaneously, a verification operation is performed on the deduced response information to check whether the deduced content conforms to the various requirements of the business adaptation rules and whether it is consistent with the actual logistics business data in the node-based data unit. Corresponding deduction verification results are generated. Referring to the judgment and feedback information in the deduction verification results, a logical generation operation of the response content is performed on the node-based data unit. According to the logistics business logic of the business adaptation rules, the effective logistics information in the node-based data unit is sorted out, and the response expression content that fits the actual node logistics business is organized to form the corresponding node response fragment.

[0055] Furthermore, the adjusted node response fragments can be node response content that adapts to the information focus and context of the original node response fragments by combining the features of the perceptual response framework with historical cases, thus meeting the information expression needs of the corresponding logistics consultation scenario. Contextualized response fragments can be response content fragments that adapt the expression style of the node response fragments that have undergone information focus adjustment, fitting the specific logistics scenario and the user's consultation understanding habits, matching the expression needs of the consultation scenario and the user's receiving preferences. Historical cases related to the features of the perceptual response framework are retrieved, and the information expression focus and core expression direction of problem handling in similar logistics consultation scenarios in historical cases are compared to... The node response segments are adjusted according to their information focus and context. Core logistics business information that matches similar historical case scenarios is selected from the node response segments, and the proportion of such information is increased while the presentation of non-core information with low relevance to the scenario is reduced. The adjusted node response segments are then adapted by adjusting their expression style. Referring to the expression forms and language styles that are highly accepted by users in similar logistics consultation scenarios in the feature association historical cases, the language expression of the adjusted node response segments is optimized to make the expression more in line with the expression requirements of specific logistics consultation scenarios and users' daily consultation understanding habits, resulting in context-based response segments.

[0056] Furthermore, the response logic can be a response content organization logic customized for a single logistics consultation request within the perception-response framework, based on the upstream and downstream relationships and flow order of logistics nodes; it can also be a response content arrangement logic that integrates the constraints of logistics business adaptation rules; or it can be a response information integration logic that combines feature association with historical cases as supplementary evidence; or it can be a response content logic association rule that fits the logistics business process. The response logic built into the perception-response framework is invoked, and the flow order of logistics nodes and business information integration requirements set in the response logic are compared. For scenario-based response segments, content logic reconstruction is performed. Scenario-based response segments corresponding to each logistics target node are integrated according to the arrangement rules of the response logic. The presentation order and information combination method of each scenario-based response segment are adjusted. Business connection descriptions between scenario-based response segments are supplemented according to the business association requirements of the response logic. Different information focuses and expression combinations are designed during the content logic reconstruction process. Multi-dimensional content logic reconstruction operations are carried out for scenario-based response segments of the same logistics consultation request, generating multiple differentiated candidate answers with differences in information presentation order, expression focus, and content combination method.

[0057] S5. Extract the intent type features from the user's inquiry intent, and filter the priority determination elements in the logistics business element set. Determine the logistics business priority of the logistics inquiry request based on the priority determination elements, and calculate the weighted score of each differentiated candidate answer based on the intent type features and the logistics business priority.

[0058] In this embodiment of the invention, the intent type feature can be a subdivided type of feature extracted from the user's logistics consultation intent, corresponding to different specific logistics consultation scenarios. It is usually a feature corresponding to various subdivided types of logistics consultation such as trajectory query, lost package compensation, timeliness complaint, address modification, and freight consultation, based on the logistics consultation scenario. This type of feature will serve as the first dimension basis for evaluating candidate answers in the intelligent logistics customer service question and answer process, and provide specific dimensional reference for the dynamic weighted calculation of candidate answers.

[0059] In detail, the core consultation intent of the logistics consultation request initiated by the user is classified by the pre-trained logistics intent classification model to obtain the corresponding user consultation intent. Then, according to the standards of the logistics consultation scenario, such as trajectory query, lost package compensation, timeliness complaint, address modification, and freight consultation, the relevant features of the matching sub-type are extracted from the classified user consultation intent to complete the extraction operation of intent type features.

[0060] In this embodiment of the invention, the priority determination element can be a key element selected from the set of logistics business elements to mark the priority level of the corresponding business of the logistics consultation request. It is usually a waybill type, delivery time requirement, and problem handling type that can reflect the urgency and processing needs of the logistics consultation business. The waybill type can include categories such as fresh food express, ordinary express, and large item logistics. The delivery time requirement can include categories such as next-day delivery, two-day delivery, and ordinary delivery. The problem handling type can include categories such as claims, complaints, and routine consultations. These elements can provide a basis for marking the priority level of the logistics consultation request and can also serve as a second dimension for evaluating candidate answers in the process of intelligent logistics customer service Q&A.

[0061] In detail, from the set of logistics business elements generated after semantic deconstruction and mapping, three key elements are selected: waybill type, delivery time requirement, and problem handling type. Waybill type corresponds to specific categories such as fresh food express, ordinary express, and large item logistics; delivery time requirement corresponds to specific categories such as next-day delivery, two-day delivery, and ordinary delivery; and problem handling type corresponds to specific categories such as claims, complaints, and routine inquiries.

[0062] In this embodiment of the invention, the logistics business priority can be determined by combining preset logistics business priority classification rules and prioritizing factors such as waybill type, delivery time requirements, and problem handling type to classify the business urgency level of logistics consultation requests. It is usually divided into three levels, corresponding to high, medium, and low business urgency levels, respectively. This classification result can provide a second dimension for evaluating candidate answers in intelligent logistics customer service Q&A and is also an important reference for dynamically weighted calculation of candidate answer scores.

[0063] In detail, in accordance with the preset logistics business priority classification rules, and combined with the three priority judgment elements of the selected waybill type, delivery time requirement, and problem handling type, the logistics consultation request is classified into three levels: Level 1, Level 2, and Level 3. Level 1 is matched with high priority, Level 2 with medium priority, and Level 3 with low priority, thereby implementing the classification of logistics business priorities.

[0064] In this embodiment of the invention, the weighted calculation score can be a score result obtained by combining the intent type features of logistics consultation with the priority level of logistics business, and by using a dynamic weighted calculation algorithm to quantify the differentiated candidate answers generated during the intelligent logistics customer service Q&A process. It is usually a specific value calculated according to a predetermined calculation formula based on the base score of intent type feature weight coefficient, intent matching degree, logistics business priority level weight coefficient, priority matching degree, and the semantic completeness of the candidate answer, as well as the additional score of scenario matching degree. This score can serve as the core basis for judging the quality of each differentiated candidate answer, and is used to subsequently select the optimal logistics customer service answer with the highest suitability.

[0065] In this embodiment of the invention, the step of calculating a weighted score for each differentiated candidate answer based on the intent type feature and the logistics business priority includes: According to the preset logistics weight configuration rules, the intention type feature and the logistics business priority are respectively assigned weight coefficients to obtain the intention feature weight coefficient and the business priority weight coefficient. The intent matching degree between each differentiated candidate answer and the intent type feature is calculated using a preset semantic matching algorithm, and the priority matching degree between each differentiated candidate answer and the logistics business priority is also calculated. Multiply the intent feature weight coefficient by the intent matching degree to obtain a first calculation result; multiply the priority weight coefficient by the priority matching degree to obtain a second calculation result; sum the first calculation result and the second calculation result to obtain a comprehensive matching calculation value. Analyze the semantic completeness base score of each differentiated candidate answer, and multiply the comprehensive value calculated by matching with the semantic completeness base score to obtain the basic weighted score; By combining the business adaptation rules with the real-time logistics status data, a corresponding scenario matching score is added to each differentiated candidate answer. The basic weighted score is then added to the scenario matching score to obtain a weighted score for each differentiated candidate answer.

[0066] In detail, the preset logistics weight configuration rules can be a rule system formulated by logistics companies based on their actual business needs and the characteristics of logistics consulting scenarios. These rules can usually be fine-tuned according to the real-time status of logistics business, and are used to assign corresponding weight coefficients to intent type features and logistics business priorities. The intent feature weight coefficient can be a numerical coefficient assigned to intent type features based on the preset logistics weight configuration rules. It is usually set differently according to different logistics consulting intent type features and is used for subsequent dynamic weighted calculations. The business priority weight coefficient can be a numerical coefficient assigned to logistics business priorities based on the preset logistics weight configuration rules. It is usually set differently according to different levels of logistics business priority and is used for subsequent dynamic weighted calculations. By referring to the preset logistics weight configuration rules, weight coefficients are assigned to intent type features to determine and assign corresponding numerical coefficients, thus forming the intent feature weight coefficients. Simultaneously, weight coefficients are assigned to logistics business priorities to determine and assign corresponding numerical coefficients, thus forming the business priority weight coefficients.

[0067] Specifically, the preset semantic matching algorithm can be a natural language processing algorithm adapted to logistics customer service Q&A scenarios. It is usually an algorithm model that can quantitatively determine the semantic fit and adaptability between different contents, providing algorithmic support for the calculation of intent matching degree and priority matching degree. The intent matching degree can be a quantitative value of the semantic fit between differentiated candidate answers and intent type features. It is usually a value between 0 and 1 and is an important parameter for dynamically weighted calculation of scores. The priority matching degree can be a quantitative value of the adaptability between differentiated candidate answers and logistics business priorities. It is usually a value between 0 and 1 and is an important parameter for dynamically weighted calculation of scores. The preset semantic matching algorithm is called to calculate the semantic fit between each differentiated candidate answer and intent type feature to obtain the corresponding intent matching degree. At the same time, the semantic adaptability between each differentiated candidate answer and logistics business priority is calculated to obtain the corresponding priority matching degree.

[0068] Furthermore, the first calculation result can be the numerical result obtained by multiplying the intent feature weight coefficient and the intent matching degree, which is usually the basic intermediate value in the dynamic weighted calculation of logistics customer service Q&A; the second calculation result can be the numerical result obtained by multiplying the priority weight coefficient and the priority matching degree, which is usually the basic intermediate value used in conjunction with the first calculation result in the dynamic weighted calculation of logistics customer service Q&A; the comprehensive matching calculation value can be the sum of the first calculation result and the second calculation result, which is usually the core intermediate value in the dynamic weighted calculation of logistics customer service Q&A that carries the basic multiplication result. The first calculation result is obtained by multiplying the intent feature weight coefficient and the intent matching degree, the second calculation result is obtained by multiplying the priority weight coefficient and the priority matching degree, and the comprehensive matching calculation value is obtained by summing the first calculation result and the second calculation result.

[0069] Furthermore, the semantic completeness base score can be a quantitative score given after evaluating the semantic completeness of each differentiated candidate answer. It is usually a value between 0 and 1 and serves as an important basic parameter in the dynamic weighted calculation of logistics customer service questions and answers. The basic weighted score can be the numerical result obtained by multiplying the comprehensive value of the matching calculation with the semantic completeness base score. It is usually the core intermediate value connecting the matching calculation and the additional score calculation of the scenario matching degree in the dynamic weighted calculation of logistics customer service questions and answers. The semantic completeness of each differentiated candidate answer is analyzed and evaluated to determine the corresponding semantic completeness base score. The basic weighted score is obtained by multiplying the comprehensive value of the matching calculation and the semantic completeness base score.

[0070] In addition, the scenario matching score can be an additional quantitative score for each differentiated candidate answer, which is obtained by combining logistics business adaptation rules and real-time logistics status data. It is usually a value between 0 and 0.3. It is an important additional parameter for optimizing the score result in the dynamic weighted calculation of logistics customer service Q&A. The scenario matching operation is carried out for each differentiated candidate answer by combining business adaptation rules and real-time logistics status data, and the corresponding scenario matching score is assigned. The basic weighted score and the scenario matching score are added together to obtain the weighted calculated score of each differentiated candidate answer.

[0071] S6. Select the differentiated candidate answer with the highest score among the weighted calculated scores as the target customer service answer.

[0072] In this embodiment of the invention, the target customer service answer can be the highest-scoring differential candidate answer determined by weighted calculation and screening from multiple differential candidate answers generated during the intelligent logistics customer service Q&A process. It is usually the logistics customer service answer that fits the user's core logistics consultation needs, meets the priority level requirements of logistics business, and has the best adaptability to the actual logistics business scenario. It is the core answer content that the intelligent logistics customer service finally outputs to the user who initiated the consultation.

[0073] In detail, the weighted scores corresponding to all differentiated candidate answers are sorted out, the numerical values ​​of each weighted score are compared, the weighted score with the highest value is selected and the differentiated candidate answer corresponding to that score is locked, the differentiated candidate answer is filtered and identified as the target customer service answer.

[0074] As can be seen, the above solution can realize customized processing of the entire process from query request parsing to target answer output, effectively improve the accuracy of intent parsing and knowledge retrieval, make the answer both real-time and logically coherent, and scientifically quantify and evaluate candidate answers to select the optimal answer that fits the core needs, which can solve the problem of low accuracy of intelligent logistics customer service Q&A.

[0075] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0076] In one embodiment, an intelligent logistics customer service Q&A device 100 is provided, which corresponds one-to-one with the intelligent logistics customer service Q&A method described in the above embodiments. For example... Figure 5 As shown, the intelligent logistics customer service Q&A device 100 includes a consultation intent and element extraction module 101, a logistics knowledge graph retrieval module 102, a perception response framework construction module 103, a candidate answer generation module 104, a candidate answer weighted score calculation module 105, and a target customer service answer filtering module 106. Detailed descriptions of each functional module are as follows: The consultation intent and element extraction module 101 is used to receive logistics consultation requests initiated by target users, perform semantic deconstruction and business element mapping processing on the logistics consultation requests, and obtain a set of user consultation intent and logistics business elements. The logistics knowledge graph retrieval module 102 is used to retrieve, according to a preset hierarchical similarity threshold filtering strategy, target logistics nodes that match the user's consultation intent and the set of logistics business elements respectively in terms of node matching, business matching, and feature matching, as well as historical cases of business adaptation rules and feature association from the preset logistics knowledge graph. The perception and response framework construction module 103 is used to construct the perception and response framework for the logistics consultation request based on the target logistics node, the business adaptation rules and the feature-associated historical cases, combined with the preset logistics scenario self-adaptation logic. The candidate answer generation module 104 is used to filter the identifying business elements with target retrieval attributes in the set of logistics business elements, retrieve the real-time logistics status data corresponding to the logistics consultation request from the preset real-time logistics business database according to the identifying business elements, integrate the real-time logistics status data into the perception response framework, and generate multiple differentiated candidate answers using the perception response framework. The candidate answer weighted score calculation module 105 is used to extract the intent type features in the user's consultation intent, filter the priority determination elements in the logistics business element set, mark the logistics business priority of the logistics consultation request according to the priority determination elements, and calculate the weighted score of each differentiated candidate answer according to the intent type features and the logistics business priority. The target customer service answer filtering module 106 is used to filter the differentiated candidate answers with the highest scores among the weighted calculated scores as the target customer service answers.

[0077] In one embodiment, the consultation intent and element extraction module 101, when performing semantic deconstruction and business element mapping processing on the logistics consultation request to obtain a set of user consultation intent and logistics business elements, is used for: Lexical and semantic analysis is performed on the logistics consultation request to obtain a logistics consultation word segmentation feature set; The logistics consultation word segmentation feature set is used to extract business entities to obtain a logistics business entity set; The logistics business entity set is subjected to feature matching analysis by calling the logistics consultation intent feature library in the pre-trained logistics intent classification model. The intent category is determined by the feature results obtained after matching to obtain the user consultation intent. Extract the intent feature dimension of the user's consultation intent and the entity feature dimension of the logistics business entity set, align the intent feature dimension and the entity feature dimension, and perform structured association and integration on the aligned intent entity feature pairs to obtain the element mapping relationship between the user's consultation intent and the logistics business entity set. Based on the element mapping relationship, the user's consultation intent and the logistics business entity set are fused together to obtain a logistics business element set.

[0078] In one embodiment, the logistics knowledge graph retrieval module 102, when executing a preset hierarchical similarity threshold filtering strategy to retrieve target logistics nodes, business matching rules, and feature-related historical cases that match the user's consultation intent and the set of logistics business elements respectively from a preset logistics knowledge graph, is used to: The user's consultation intent is decomposed into contextual features to obtain an intent scenario feature set. The logistics business element set is decomposed into business features to obtain a business attribute feature set. The intent scenario feature set and the business attribute feature set are fused to obtain a target retrieval feature set. Using the retrieval dimensions corresponding to the three-level structure in the preset logistics knowledge graph, hierarchical feature matching is performed on the target retrieval feature set to obtain the matching similarity value of each level. The matching similarity values ​​of each level are compared with the preset level similarity threshold to filter out candidate content whose matching similarity values ​​are greater than or equal to the corresponding level similarity threshold. The candidate content is mapped to process-related nodes to obtain a node-related candidate set; Identify the matching relationship between the node association candidate set and the logistics business element set, and select target logistics nodes, business adaptation rules and feature association historical cases from the node association candidate set according to the matching relationship.

[0079] In one embodiment, the perception-response framework construction module 103, when executing the construction of a perception-response framework for the logistics consultation request based on the target logistics node, the business adaptation rules, and the feature-associated historical cases, combined with preset logistics scenario self-adaptation logic, is used for: The target logistics node is analyzed to identify the logistics node links; Logical constraints are applied to the logistics node links according to the business adaptation rules to obtain the logistics node logical links, and logical nodes are configured for the logistics node logical links according to the preset logistics scenario self-adaptation logic. The historical cases associated with the aforementioned features are decomposed into logistics scenario dimensions to obtain supplementary case-based evidence; By integrating the aforementioned case-based supplementary criteria into each logical node of the logistics node logical link, a perception-response framework is obtained.

[0080] In one embodiment, the candidate answer generation module 104, when performing the filtering of identifying business elements with target retrieval attributes in the set of logistics business elements, is used to: Based on the preset retrieval attribute dimensions, the feature information of each business element in the set of logistics business elements is classified and decomposed to obtain the retrieval attribute feature set corresponding to each business element; Extract the dimensional feature requirements corresponding to the target retrieval attributes, and quantify and decompose the dimensional feature requirements dimension by dimension according to the retrieval attribute dimensions to obtain the quantitative evaluation index corresponding to each retrieval attribute dimension, and assign corresponding index weights to each quantitative evaluation index. Using the quantitative evaluation indicators and the corresponding indicator weights, the dimensional feature matching calculation is performed on the retrieval attribute feature set corresponding to each business element to obtain the matching score of each retrieval attribute dimension. The matching scores of each search attribute dimension are weighted and summed according to the corresponding indicator weights to obtain the dimension fit value of each business element. All business elements are sorted in descending order of their dimensional fit value to obtain an adaptation ranking result. The business elements in the preset first-place interval of the adaptation ranking result are identified as identifiable business elements with target retrieval attributes.

[0081] In one embodiment, the candidate answer generation module 104, when generating multiple differentiated candidate answers using the perceptual response framework, is further configured to: The target logistics node of the perception and response framework is used to analyze the logistics business dimension corresponding to the fused data. Based on the logistics business dimension, the fused data is decomposed into node-dimensional data units. The business adaptation rules of the perception and response framework are used to deduce and verify the response information of the nodeized data unit. Based on the deduction and verification results, the response content of the nodeized data unit is logically generated to obtain the node response fragment. By combining the features of the perception and response framework with historical cases, the information focus of the node response fragment is adjusted in a contextualized manner, and the expression style of the adjusted node response fragment is adapted to obtain a contextualized response fragment. Based on the built-in response logic of the perception response framework, the content logic of the scenario-based response fragment is reconstructed to generate multiple differentiated candidate answers.

[0082] In one embodiment, the candidate answer weighted score calculation module 105, when performing the weighted score calculation for each differentiated candidate answer based on the intent type characteristics and the logistics business priority, is used to: According to the preset logistics weight configuration rules, the intention type feature and the logistics business priority are respectively assigned weight coefficients to obtain the intention feature weight coefficient and the business priority weight coefficient. The intent matching degree between each differentiated candidate answer and the intent type feature is calculated using a preset semantic matching algorithm, and the priority matching degree between each differentiated candidate answer and the logistics business priority is also calculated. Multiply the intent feature weight coefficient by the intent matching degree to obtain a first calculation result; multiply the priority weight coefficient by the priority matching degree to obtain a second calculation result; sum the first calculation result and the second calculation result to obtain a comprehensive matching calculation value. Analyze the semantic completeness base score of each differentiated candidate answer, and multiply the comprehensive value calculated by matching with the semantic completeness base score to obtain the basic weighted score; By combining the business adaptation rules with the real-time logistics status data, a corresponding scenario matching score is added to each differentiated candidate answer. The basic weighted score is then added to the scenario matching score to obtain a weighted score for each differentiated candidate answer.

[0083] This invention provides an intelligent logistics customer service question-and-answer device that can realize customized processing of the entire process from parsing the consultation request to outputting the target answer. It effectively improves the accuracy of intent parsing and knowledge retrieval, makes the answer both real-time and logically coherent, and can scientifically quantify and evaluate candidate answers to select the optimal answer that fits the core needs. This can solve the problem of low accuracy in intelligent logistics customer service question-and-answer.

[0084] Specific limitations regarding the intelligent logistics customer service Q&A device can be found in the limitations of the intelligent logistics customer service Q&A method described above, and will not be repeated here. Each module in the aforementioned intelligent logistics customer service Q&A device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0085] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a smart logistics customer service question-and-answer method on the server side.

[0086] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 7 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the client-side functions or steps of an intelligent logistics customer service Q&A method.

[0087] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Receive logistics consultation requests initiated by target users, perform semantic deconstruction and business element mapping on the logistics consultation requests, and obtain a set of user consultation intent and logistics business elements. Based on a preset hierarchical similarity threshold filtering strategy, the system retrieves target logistics nodes, business matching rules, and feature-related historical cases from a preset logistics knowledge graph that match the user's consultation intent and the set of logistics business elements, respectively, based on node matching, business matching, and feature matching. Based on the target logistics node, the business adaptation rules, and the historical cases associated with the features, a perception and response framework for the logistics consultation request is constructed in conjunction with the preset logistics scenario self-adaptation logic. The system filters out identifying business elements with target retrieval attributes from the set of logistics business elements, retrieves real-time logistics status data corresponding to the logistics inquiry request from the preset real-time logistics business database based on the identifying business elements, integrates the real-time logistics status data into the perception and response framework, and uses the perception and response framework to generate multiple differentiated candidate answers. Extract the intent type features from the user's inquiry intent, and filter the priority determination elements in the logistics business element set. Determine the logistics business priority of the logistics inquiry request based on the priority determination elements. Calculate the weighted score of each differentiated candidate answer based on the intent type features and the logistics business priority. The candidate answer with the highest weighted score is selected as the target customer service answer.

[0088] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Receive logistics consultation requests initiated by target users, perform semantic deconstruction and business element mapping on the logistics consultation requests, and obtain a set of user consultation intent and logistics business elements. Based on a preset hierarchical similarity threshold filtering strategy, the system retrieves target logistics nodes, business matching rules, and feature-related historical cases from a preset logistics knowledge graph that match the user's consultation intent and the set of logistics business elements, respectively, based on node matching, business matching, and feature matching. Based on the target logistics node, the business adaptation rules, and the historical cases associated with the features, a perception and response framework for the logistics consultation request is constructed in conjunction with the preset logistics scenario self-adaptation logic. The system filters out identifying business elements with target retrieval attributes from the set of logistics business elements, retrieves real-time logistics status data corresponding to the logistics inquiry request from the preset real-time logistics business database based on the identifying business elements, integrates the real-time logistics status data into the perception and response framework, and uses the perception and response framework to generate multiple differentiated candidate answers. Extract the intent type features from the user's inquiry intent, and filter the priority determination elements in the logistics business element set. Determine the logistics business priority of the logistics inquiry request based on the priority determination elements. Calculate the weighted score of each differentiated candidate answer based on the intent type features and the logistics business priority. The candidate answer with the highest weighted score is selected as the target customer service answer.

[0089] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0090] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0091] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0092] It should be noted that if any software tools or components not belonging to our company appear in the embodiments of this application, they are merely for illustrative purposes and do not represent actual use.

[0093] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A smart logistics customer service question-and-answer method, characterized in that, include: Receive logistics consultation requests initiated by target users, perform semantic deconstruction and business element mapping on the logistics consultation requests, and obtain a set of user consultation intent and logistics business elements. Based on a preset hierarchical similarity threshold filtering strategy, the system retrieves target logistics nodes, business matching rules, and feature-related historical cases from a preset logistics knowledge graph that match the user's consultation intent and the set of logistics business elements, respectively, based on node matching, business matching, and feature matching. Based on the target logistics node, the business adaptation rules, and the historical cases associated with the features, a perception and response framework for the logistics consultation request is constructed in conjunction with the preset logistics scenario self-adaptation logic. The system filters out identifying business elements with target retrieval attributes from the set of logistics business elements, retrieves real-time logistics status data corresponding to the logistics inquiry request from the preset real-time logistics business database based on the identifying business elements, integrates the real-time logistics status data into the perception and response framework, and uses the perception and response framework to generate multiple differentiated candidate answers. Extract the intent type features from the user's inquiry intent, and filter the priority determination elements in the logistics business element set. Determine the logistics business priority of the logistics inquiry request based on the priority determination elements. Calculate the weighted score of each differentiated candidate answer based on the intent type features and the logistics business priority. The candidate answer with the highest weighted score is selected as the target customer service answer.

2. The intelligent logistics customer service Q&A method as described in claim 1, characterized in that, The semantic deconstruction and business element mapping of the logistics consultation request yields a set of user consultation intent and logistics business elements, including: Lexical and semantic analysis is performed on the logistics consultation request to obtain a logistics consultation word segmentation feature set; The logistics consultation word segmentation feature set is used to extract business entities to obtain a logistics business entity set; The logistics business entity set is subjected to feature matching analysis by calling the logistics consultation intent feature library in the pre-trained logistics intent classification model. The intent category is determined by the feature results obtained after matching to obtain the user consultation intent. Extract the intent feature dimension of the user's consultation intent and the entity feature dimension of the logistics business entity set, align the intent feature dimension and the entity feature dimension, and perform structured association and integration on the aligned intent entity feature pairs to obtain the element mapping relationship between the user's consultation intent and the logistics business entity set. Based on the element mapping relationship, the user's consultation intent and the logistics business entity set are fused together to obtain a logistics business element set.

3. The intelligent logistics customer service Q&A method as described in claim 1, characterized in that, The step of using a preset hierarchical similarity threshold filtering strategy to retrieve target logistics nodes, business matching rules, and feature association historical cases from a preset logistics knowledge graph that match the user's inquiry intent and the set of logistics business elements, respectively, including: The user's consultation intent is decomposed into contextual features to obtain an intent scenario feature set. The logistics business element set is decomposed into business features to obtain a business attribute feature set. The intent scenario feature set and the business attribute feature set are fused to obtain a target retrieval feature set. Using the retrieval dimensions corresponding to the three-level structure in the preset logistics knowledge graph, hierarchical feature matching is performed on the target retrieval feature set to obtain the matching similarity value of each level. The matching similarity values ​​of each level are compared with the preset level similarity threshold to filter out candidate content whose matching similarity values ​​are greater than or equal to the corresponding level similarity threshold. The candidate content is mapped to process-related nodes to obtain a node-related candidate set; Identify the matching relationship between the node association candidate set and the logistics business element set, and select target logistics nodes, business adaptation rules and feature association historical cases from the node association candidate set according to the matching relationship.

4. The intelligent logistics customer service Q&A method as described in claim 1, characterized in that, The process of constructing a perception and response framework for the logistics inquiry request based on the target logistics node, the business adaptation rules, and the historical cases associated with the features, combined with a preset logistics scenario self-adaptation logic, includes: The target logistics node is analyzed to identify the logistics node links; Logical constraints are applied to the logistics node links according to the business adaptation rules to obtain the logistics node logical links, and logical nodes are configured for the logistics node logical links according to the preset logistics scenario self-adaptation logic. The historical cases associated with the aforementioned features are decomposed into logistics scenario dimensions to obtain supplementary case-based evidence; By integrating the aforementioned case-based supplementary criteria into each logical node of the logistics node logical link, a perception-response framework is obtained.

5. The intelligent logistics customer service Q&A method as described in claim 1, characterized in that, The process of filtering the identifying business elements with target retrieval attributes from the set of logistics business elements includes: Based on the preset retrieval attribute dimensions, the feature information of each business element in the set of logistics business elements is classified and decomposed to obtain the retrieval attribute feature set corresponding to each business element; Extract the dimensional feature requirements corresponding to the target retrieval attributes, and quantify and decompose the dimensional feature requirements dimension by dimension according to the retrieval attribute dimensions to obtain the quantitative evaluation index corresponding to each retrieval attribute dimension, and assign corresponding index weights to each quantitative evaluation index. Using the quantitative evaluation indicators and the corresponding indicator weights, the dimensional feature matching calculation is performed on the retrieval attribute feature set corresponding to each business element to obtain the matching score of each retrieval attribute dimension. The matching scores of each search attribute dimension are weighted and summed according to the corresponding indicator weights to obtain the dimension fit value of each business element. All business elements are sorted in descending order of their dimensional fit value to obtain an adaptation ranking result. The business elements in the preset first-place interval of the adaptation ranking result are identified as identifiable business elements with target retrieval attributes.

6. The intelligent logistics customer service Q&A method as described in claim 1, characterized in that, The generation of multiple differentiated candidate answers using the perceptual response framework includes: The target logistics node of the perception and response framework is used to analyze the logistics business dimension corresponding to the fused data. Based on the logistics business dimension, the fused data is decomposed into node-dimensional data units. The business adaptation rules of the perception and response framework are used to deduce and verify the response information of the nodeized data unit. Based on the deduction and verification results, the response content of the nodeized data unit is logically generated to obtain the node response fragment. By combining the features of the perception and response framework with historical cases, the information focus of the node response fragment is adjusted in a contextualized manner, and the expression style of the adjusted node response fragment is adapted to obtain a contextualized response fragment. Based on the built-in response logic of the perception response framework, the content logic of the scenario-based response fragment is reconstructed to generate multiple differentiated candidate answers.

7. The intelligent logistics customer service question-and-answer method as described in claim 1, characterized in that, The step of calculating a weighted score for each differentiated candidate answer based on the intent type characteristics and the logistics business priority includes: According to the preset logistics weight configuration rules, the intention type feature and the logistics business priority are respectively assigned weight coefficients to obtain the intention feature weight coefficient and the business priority weight coefficient. The intent matching degree between each differentiated candidate answer and the intent type feature is calculated using a preset semantic matching algorithm, and the priority matching degree between each differentiated candidate answer and the logistics business priority is also calculated. Multiply the intent feature weight coefficient by the intent matching degree to obtain a first calculation result; multiply the priority weight coefficient by the priority matching degree to obtain a second calculation result; sum the first calculation result and the second calculation result to obtain a comprehensive matching calculation value. Analyze the semantic completeness base score of each differentiated candidate answer, and multiply the comprehensive value calculated by matching with the semantic completeness base score to obtain the basic weighted score; By combining the business adaptation rules with the real-time logistics status data, a corresponding scenario matching score is added to each differentiated candidate answer. The basic weighted score is then added to the scenario matching score to obtain a weighted score for each differentiated candidate answer.

8. An intelligent logistics customer service Q&A device, characterized in that, include: The consultation intent and element extraction module is used to receive logistics consultation requests initiated by target users, perform semantic deconstruction and business element mapping processing on the logistics consultation requests, and obtain a set of user consultation intent and logistics business elements. The logistics knowledge graph retrieval module is used to retrieve, according to a preset hierarchical similarity threshold filtering strategy, target logistics nodes that match the user's consultation intent and the set of logistics business elements respectively in terms of node matching, business matching, and feature matching, as well as historical cases of business adaptation rules and feature association from the preset logistics knowledge graph. The perception and response framework construction module is used to construct the perception and response framework for the logistics consultation request based on the target logistics node, the business adaptation rules and the historical cases associated with the features, combined with the preset logistics scenario self-adaptation logic. The candidate answer generation module is used to filter out the identifying business elements with target retrieval attributes in the set of logistics business elements, retrieve the real-time logistics status data corresponding to the logistics consultation request from the preset real-time logistics business database according to the identifying business elements, integrate the real-time logistics status data into the perception response framework, and use the perception response framework to generate multiple differentiated candidate answers. The candidate answer weighted score calculation module is used to extract the intent type features from the user's consultation intent, filter the priority determination elements in the logistics business element set, mark the logistics business priority of the logistics consultation request according to the priority determination elements, and calculate the weighted score of each differentiated candidate answer according to the intent type features and the logistics business priority. The target customer service response filtering module is used to filter the differentiated candidate responses with the highest scores among the weighted calculated scores as the target customer service responses.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent logistics customer service Q&A method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the intelligent logistics customer service question-and-answer method as described in any one of claims 1 to 7.

Images