A logistics full-link visualized collaboration method, system, medium and product

Abstract

The application provides a logistics full-link visual collaboration method, system, medium and product, relates to the field of logistics informatization, and the method comprises the following steps: extracting operation intention labels and logistics key entities from user logistics collaboration requirements; matching an interactive interface according to the intention labels, combining the key entities and historical operation data to construct a candidate information set for each to-be-filled field; evaluating the transmission influence of each candidate information based on a logistics data dependency topology, selecting a first target candidate information with the minimum restriction on surrounding fields to perform preliminary filling; when the user modifies a field, deriving a constraint boundary along the topology to locate an affected field, extracting a non-conflict attribute value of the affected field as a reserved attribute feature, and then selecting a second target candidate information from a corresponding candidate set, which satisfies the constraint boundary and has the highest comprehensive matching degree with the reserved attribute feature and the intention label, to perform local replacement. The method can improve the form entry and interactive operation efficiency in the logistics collaboration scenario.

Description

Technical Field

[0001] This application relates to the field of logistics information technology, and in particular to a visual collaborative method, system, medium and product for the entire logistics chain. Background Technology

[0002] With the deep integration of the logistics supply chain, modern logistics operations generally involve a large number of multi-stage collaborations. To ensure the smooth progress of these multi-party operations, existing digital business platforms generally provide visual interactive interfaces for collaborators to enter logistics orders, configure nodes, and fill in various business parameters.

[0003] In related technologies, rule-tree-based form cascading control is commonly used to handle changes in collaborative data entry. In this approach, the system internally constructs a robust set of data cascading verification rules. When a user modifies a key upstream business parameter on the user interface, the system automatically resets all associated downstream fields to their initial blank state or system default values, based on preset mapping rules. This ensures data compliance throughout the logistics form's subsequent flow and prevents logical conflicts between upstream and downstream parameters.

[0004] However, facing the new demands of sophisticated and highly interconnected logistics collaboration, an order often involves complex nested attributes. When changes on-site necessitate user modifications to certain parameters, the aforementioned cascading reset mechanism, while ensuring no conflicts, indiscriminately clears a large amount of configured valid data in downstream fields. This necessitates extensive re-verification and secondary data entry for the cleared areas by business collaboration personnel, reducing operational efficiency. Furthermore, this cascading reset mechanism triggers extensive re-rendering of downstream child nodes in the front-end DOM tree and server-side re-validation requests for all downstream fields at the system level, causing browser rendering thread blocking and wasting server I / O resources, further degrading system response performance in high-concurrency collaboration scenarios. Summary of the Invention

[0005] This application provides a visual collaboration method, system, medium, and product for the entire logistics chain, which can improve the efficiency of form entry and interactive operation in logistics collaboration scenarios.

[0006] Firstly, this application provides a method for visualizing and collaborating across the entire logistics chain, the method comprising:

[0007] Semantic parsing is performed on the logistics collaboration requirements input by the user to extract operation intent tags and key logistics entities, with each operation intent tag corresponding to an intent representation vector;

[0008] Match the logistics human-computer interaction interface to be rendered according to the operation intention tag, and construct a corresponding candidate information set for each field to be filled in the logistics human-computer interaction interface according to the logistics key entity and the pre-acquired logistics historical operation data.

[0009] Based on the pre-stored logistics data, each candidate information in the candidate information set is evaluated and calculated to determine the first target candidate information that has the least impact on the change of other fields to be filled. The first target candidate information is used to initially fill the fields to be filled in the logistics human-computer interaction interface and to display the initial interface assembly.

[0010] In response to a user's modification operation on the target field to be filled through the logistics human-computer interaction interface, the derived constraint boundary is generated by extracting the new content to be filled corresponding to the modification operation based on the logistics data dependency topology, and the affected field that conflicts with the derived constraint boundary is located.

[0011] Based on the attribute dimension values ​​in the initial filling content of the affected field that do not conflict with the derived constraint boundary, retain attribute features are constructed.

[0012] Candidate information that satisfies the derived constraint boundary is selected from the candidate information set corresponding to the affected field. The feature distance of the selected candidate information is calculated based on the preserved attribute features and the intent representation vector. The candidate information with the smallest feature distance is selected as the second target candidate information with the highest comprehensive matching degree, so as to drive the rendering node corresponding to the affected field to perform local data update and replacement display.

[0013] By adopting the above technical solution, when the system receives a user's partial modification to a field, it extracts derived constraint boundaries based on data dependency topology to accurately locate the affected range and extracts non-conflicting attributes as retained features. Subsequently, it performs a partial replacement on the affected field based on the user's modification intent. This method replaces the traditional cascading full clear mechanism with a partial update, allowing related data without logical conflicts to be retained, reducing the workload required for users to re-verify and re-fill a large number of blank downstream forms after modifying upstream parameters.

[0014] In some embodiments, the step of constructing a corresponding candidate information set for each field to be filled in the logistics human-computer interaction interface based on the key logistics entity and the pre-acquired historical logistics operation data specifically includes:

[0015] Based on the key logistics entities, an initial candidate set is obtained by performing an association search on the historical logistics operation data.

[0016] Acquire real-time status warning data of the current logistics network environment, and remove candidate information with abnormal attribute features that match the real-time status warning data from the initial candidate set to obtain a usable candidate set;

[0017] Each candidate information in the available candidate set is converted into a business feature vector, and the feature distance between different candidate information within the available candidate set is calculated.

[0018] Based on the feature distance, the candidate information in the available candidate set is sampled to generate the candidate information set that meets the preset capacity upper limit threshold. The feature distance between any two candidate information in the candidate information set is greater than the preset differentiation threshold.

[0019] By adopting the above technical solution, the system not only eliminates unusable data in the current network environment, but also accurately filters redundant and highly similar options. As a result, only a small number of distinct terms are presented in the drop-down or pop-up options on the front-end interface, allowing users to quickly select the desired content without having to tediously search through long lists.

[0020] In some embodiments, the step of evaluating and calculating each candidate information in the candidate information set based on the pre-stored logistics data dependency topology to determine the first target candidate information that has the least impact on the change magnitude of other fields to be filled specifically includes:

[0021] Each candidate information in the candidate information set is sequentially used as a trigger node and input into the logistics data dependency topology to simulate constraint transmission, thereby extracting the cascading impact path that affects other fields to be filled.

[0022] For each candidate information in the candidate information set, based on the cascading influence path, the number of options in the candidate information set that fail due to constraint conflicts for other fields to be filled affected by the candidate information is counted, and the number of options is summarized to calculate the space compression cost corresponding to the candidate information.

[0023] The candidate information with the lowest spatial compression cost is determined as the first target candidate information.

[0024] By employing the above technical solution, the system sequentially substitutes candidate information into the data dependency topology for constraint simulation and propagation. This counts the number of invalid options caused by the candidate information to surrounding unfilled nodes, calculates the space compression cost, and prioritizes the option with the lowest cost. During the initial automatic filling of the interface, this method prioritizes the feature parameters with the most lenient restrictions on other related form items, ensuring that users still have a wide range of valid dropdown options when filling in other surrounding fields, thus reducing the rate of repeated modifications caused by limited input.

[0025] In some embodiments, the step of extracting the derived constraint boundary generated from the newly filled content corresponding to the modification operation based on the logistics data dependency topology specifically includes:

[0026] Obtain the new fill content input to the target field to be filled corresponding to the modification operation;

[0027] The newly filled content is compared with the original preliminary filled content of the target field to be filled to extract the logistics attribute bias that has changed.

[0028] Using the target field to be filled as the baseline node, the downstream nodes that have a causal relationship with the baseline node are traversed in the logistics data dependency topology, and the business constraint rules attached to the downstream nodes are obtained. The business constraint rules include the business attribute dimensions corresponding to the downstream nodes.

[0029] The logistics attribute bias is substituted into the business constraint rules to perform influence range extrapolation, generating the derived constraint boundary composed of multi-dimensional attribute constraints. The derived constraint boundary is used to limit the range of legal candidate data allowed for the downstream node after being affected by the modification operation. Each dimension in the multi-dimensional attribute constraints corresponds to an attribute dimension of the affected field.

[0030] By adopting the above technical solution, the system compares the attribute biases before and after the modification, extends the data dependency topology to downstream nodes, and transforms the derived constraint boundaries of multi-dimensional attribute dimensions according to the attached business rules. The cascading effects are concretized as the accurate and legal available data value range of each affected field, thereby defining the hard filtering boundaries of candidate information that can be filled in each downstream related field after this modification.

[0031] In some embodiments, the step of filtering candidate information that satisfies the derived constraint boundary from the candidate information set corresponding to the affected field, calculating the feature distance of the filtered candidate information based on the preserved attribute features and the intent representation vector, and selecting the candidate information with the smallest feature distance as the second target candidate information with the highest comprehensive matching degree to drive the rendering node corresponding to the affected field to perform local data update and replacement display specifically includes:

[0032] Candidate information that does not conflict with the derived constraint boundary is extracted from the candidate information set corresponding to the affected field to obtain a verification candidate set;

[0033] Based on the node-level characteristics of the logistics data dependency topology, the values ​​of each non-conflicting attribute dimension contained in the retained attribute characteristics are transformed into multi-dimensional vectorization to generate a baseline state vector.

[0034] Each candidate information in the verification candidate set is simulated and substituted into the node position corresponding to the affected field in the logistics data dependency topology, and the candidate connected subgraph activated by the transmission of each candidate information along the logistics data dependency topology is extracted.

[0035] The node attribute features of each candidate connected subgraph are encoded according to the baseline state vector to obtain the candidate state vector;

[0036] Calculate the Euclidean distance between the baseline state vector and each of the candidate state vectors, and select the candidate information corresponding to the smallest Euclidean distance as the second target candidate information, so as to drive the rendering node corresponding to the affected field to perform local data update and replacement display.

[0037] By adopting the above technical solution, the system substitutes candidate information that satisfies the constraint boundaries into the data dependency topology and extracts the corresponding candidate connected subgraphs. Then, it uses the baseline state vector to perform graph encoding and selects the candidate information with the smallest Euclidean distance to perform the interface replacement operation. This method not only considers the legal connectivity of single-point parameters but also finds the option that is closest to the filling attribute of the original affected area. This ensures that the new content automatically replaced on the interface by the system best matches the user's previously configured parameter type preferences, reducing the frequency of manual reselection.

[0038] In some embodiments, the step of performing multi-dimensional vectorization transformation on the values ​​of each non-conflicting attribute dimension included in the retained attribute features based on the node-level characteristics of the logistics data dependency topology to generate a baseline state vector specifically includes:

[0039] Extract the deviation direction between the new filling content and the initial filling content corresponding to the modification operation in the feature semantic space, and generate an intention drift vector by combining the logistics attribute bias.

[0040] The intent representation vector and the business attribute dimension of the business constraint rule are jointly mapped to the same source representation space to generate an initial intent reference vector, and the feature distance between the intent drift vector and the initial intent reference vector is calculated to obtain the semantic discrete bias.

[0041] When the semantic discrete bias is detected to be less than the preset redirection threshold, the baseline state vector is directly generated based on the retained attribute features;

[0042] When the semantic discrete bias is detected to exceed the preset redirection threshold, the values ​​of each non-conflicting attribute dimension contained in the retained attribute feature are used to construct an initial attribute matrix.

[0043] Obtain the adjacent association edge weights of the affected fields in the logistics data dependency topology, and use the adjacent association edge weights to weight the indicators in the initial attribute matrix to generate a local feature weighted matrix;

[0044] Based on the node-level characteristics of the logistics data dependency topology, the matrix parameters contained in the local feature weighting matrix are encoded using multidimensional vectorization to obtain the baseline state vector.

[0045] By adopting the above technical solution, after quantifying the semantic offset caused by local adjustments, if the deviation exceeds the preset threshold for redirection, the system uses the weights of topological adjacency-related edges to complete local weighted alignment to generate a realignment matrix, which is ultimately multidimensionally vectorized into a baseline state vector. When the user's core intent in modifying the form changes significantly, the system can dynamically correct the matching baseline through this geometric correction method, ensuring that the candidate options automatically replacing the user on the interface closely follow the user's sudden change in true filling intent, making the displayed updated items accurately match the user's actual expectations.

[0046] In some embodiments, after the steps of filtering candidate information that satisfies the derived constraint boundary from the candidate information set corresponding to the affected field, calculating the feature distance of the filtered candidate information based on the preserved attribute features and the intent representation vector, and selecting the candidate information with the smallest feature distance as the second target candidate information with the highest comprehensive matching degree to drive the rendering node corresponding to the affected field to perform local data update and replacement display, the method further includes:

[0047] Extract the feature deviation between the initial filling content corresponding to the affected field and the second target candidate information, and perform a weighted calculation on the feature deviation with a preset business weight matrix to obtain the verification priority parameter. The business weight matrix is ​​composed of the business importance coefficients of each node in the logistics data dependency topology.

[0048] When the verification priority parameter is greater than the preset threshold, the directed transmission path from the target field to be filled to the affected field in the logistics data dependency topology is mapped to the screen coordinate system of the logistics interaction interface to generate a page linkage guidance trajectory.

[0049] The floating interface component is rendered based on the page linkage guide trajectory, and the initial filling content and the second target candidate information are output in parallel in the floating interface component to generate a change comparison view;

[0050] In response to the update confirmation interaction command for the change comparison view within the scope of the page linkage guidance trajectory, the affected fields are updated and confirmed.

[0051] By adopting the above technical solution, after the system completes the replacement scheme selection, it calculates the verification priority based on the feature deviation and the business weight of the node. When it is determined that the change has a significant impact, it directly generates a guiding trajectory from the modification point to the affected point and a juxtaposed comparison view on the interface, awaiting instruction interaction. This method transforms the implicit automatic change action in the background into a visible connection and comparison pop-up window on the front end, clearly informing users with visual animations which high-risk fields have undergone related numerical changes due to the operation, preventing users from making incorrect entries or false reports due to not noticing related changes.

[0052] Secondly, this application provides a logistics end-to-end visual collaborative system, the system comprising: one or more processors and a memory;

[0053] The memory is coupled to the one or more processors. The memory is used to store computer program code, which includes computer instructions. The one or more processors call the computer instructions so that the system can implement the logistics end-to-end visualization and collaboration method provided in the above embodiments, which will not be described in detail here.

[0054] Thirdly, this application provides a computer-readable storage medium including instructions that, when executed on a logistics end-to-end visualization and collaboration system, enable the system to implement a logistics end-to-end visualization and collaboration method provided in the above embodiments, which will not be elaborated further here.

[0055] Fourthly, this application provides a computer program product, including a computer program that, when run on a logistics end-to-end visualization and collaboration system, enables the system to implement a logistics end-to-end visualization and collaboration method provided in the above embodiments, which will not be elaborated here.

[0056] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0057] 1. The system extracts the bias generated by the current modification operation using a differential algorithm and extrapolates downstream along the data dependency topology to generate derived constraint boundaries to define the range of legal values. Within the constrained area, the system extracts and retains the attribute features of the original filled data that do not conflict with the rules, and combines these features to select the most matching candidate options to locally replace the affected fields. This topology-based local update mechanism ensures the legality of multidimensional form data flow while avoiding the direct forced clearing of downstream conflict-free data, directly reducing the workload of operators re-checking related items and performing secondary manual entry.

[0058] 2. During the initial pre-filling of candidate data, the system sequentially substitutes each candidate option into the data dependency topology to simulate constraint propagation. It then calculates the space compression cost by counting the number of surrounding related field options that become invalid, and prioritizes the option with the lowest cost for initial filling. This method minimizes the restrictions on surrounding fields during the initial pre-selection process, ensuring users still have sufficient available dropdown options when filling in subsequent parameters. Furthermore, when user parameter modifications cause significant discrete shifts in the overall data intent, the system uses topological adjacency edge weights to perform local weighted alignment. Homomorphic nodes are extracted using graph structure and Euclidean distance for replacement display, ensuring that the automatically refreshed related items instantly match the user's modified parameter selection preferences.

[0059] 3. When outputting the candidate list to the front-end interface, the system calculates the feature distance between different options in the candidate set and selects a small candidate set based on the principle of maximizing the sum of distances for high inclusiveness. This sampling and filtering process can eliminate redundant entries with highly similar attributes, allowing users to quickly select their target from a small number of candidates with obvious feature differences without extensive scrolling in the form controls. After the form is automatically replaced, the system calculates the verification priority based on feature deviation and business importance. When a preset threshold is reached, the system automatically converts the data flow path into a dynamic linkage guide trajectory on the UI and pops up a comparison view of the old and new data. This allows users to intuitively observe the changes in other linked field data caused by their own operations, preventing erroneous submissions caused by failure to notice backend changes. Attached Figure Description

[0060] Figure 1 This is a flowchart illustrating a logistics end-to-end visual collaboration method according to an embodiment of this application;

[0061] Figure 2 This is another flowchart illustrating a logistics end-to-end visual collaboration method in the embodiments of this application;

[0062] Figure 3 This is a schematic diagram of the physical device structure of a logistics end-to-end visual collaborative system in this application embodiment. Detailed Implementation

[0063] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0064] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0065] The key technological contribution of this application lies in the rational scheduling of internal computer system resources and the fine-grained update control of interface rendering nodes. Compared to the redrawing of the entire DOM subtree and the full server-side verification I / O overhead caused by traditional cascading clear control, this solution obtains the minimum affected range through topology deduction based on logistics data, and then performs local data updates and replacements only on the affected rendering nodes. This significantly reduces the amount of front-end layer redrawing and the number of server-side verification requests, improving the underlying operating efficiency of the system and representing an improvement in the internal technical performance of the computer system.

[0066] To facilitate understanding of the core technical concept of this application, the execution environment and application scenarios on which the method of this application is based will be described in detail first.

[0067] This proposed solution is primarily implemented in digital logistics collaboration systems (such as SaaS-based logistics management platforms). Its system architecture typically includes a front-end visual interactive page and a back-end business logic processing engine. This solution aims to address the technical challenges arising from frequent, broad-based cascading interventions between the front-end and back-end during the collaborative data entry process for complex logistics orders. These challenges include high system rendering overhead and repetitive manual data entry.

[0068] In related technologies, logistics forms have strict upstream and downstream constraints (for example, if an upstream option for a specific dimension is selected, the alternative values ​​for downstream fields will be restricted and locked). Existing form cascading control technology is a "one-size-fits-all" state reset: when the upstream basic parameters change, the backend system directly sends a full clear command, forcibly destroying and resetting the downstream field content of all related subtree nodes in the frontend to the initial blank. This approach not only causes a large-scale invalid redraw of the frontend DOM (Document Object Model) tree and a waste of server data verification I / O at the computer level, but also results in the loss of a large amount of downstream business data that was not only legitimate but also required a high degree of manual consideration to determine (such as temperature and humidity ranges, unloading requirements) for business operators.

[0069] To address the aforementioned shortcomings and optimize front-end interaction, this application provides a method, system, medium, and product for end-to-end logistics visualization and collaboration. The core technical concepts of this application are described below using a specific application scenario:

[0070] Phase 1: The system receives a simple business request text input by the front-end salesperson, such as "Next Monday, a batch of fresh produce will be shipped from Shanghai Cold Chain Warehouse 1 to Chengdu." The system performs back-end semantic analysis to locate the user's operational intent and retrieves the corresponding multimodal transport order interface template. It's important to note that at this stage, the system does not randomly pre-fill the interface; instead, it utilizes a pre-built "data dependency topology network" in the backend to calculate candidate options. The system prioritizes automatically pre-filling values ​​that minimize the "space compression cost" for other downstream nodes (i.e., the most inclusive and least likely to deplete the downstream dropdown list). At this point, the front-end displays an initial confirmation form with a very high degree of pre-filling completeness to the salesperson.

[0071] Phase Two: When a salesperson reviews the initial form, they may need to modify an upstream source control point due to unforeseen circumstances. For example, they might manually change the "Transportation Method: Cold Chain Full Truckload" initially entered in the system to "Less-than-Truckload (LTL)". Upon detecting this front-end modification, the back-end system does not send a standard full form clear command. Instead, it uses this node as a baseline and initiates a state derivation and deduction downstream along the "data dependency topology".

[0072] Phase Three: When the system reaches the downstream related fields through network inference, it finds that "Less-than-Truckload (LTL)" conflicts with the initially filled "Carrier: A Cold Chain Direct Line"; however, it does not conflict with the initially filled downstream field "Temperature Control Requirement: 0-4 degrees Celsius" (this parameter is still valid and usable). Therefore, this application's system extracts valid data such as "0-4 degrees Celsius" as "reserved attribute features". To solve the computational problem that specific parameter values ​​and macro-level user intentions cannot be directly compared in the feature dimension, the system maps the above-mentioned "reserved attribute features" and the "representation vector corresponding to user operation intentions" extracted in Phase One to the same set of homogeneous coding feature spaces, synthesizing a verification benchmark with high-dimensional comprehensive features. Based on this verification benchmark, the system calculates and searches the database for the target closest to the new feature (i.e., a new carrier B that both performs LTL and fully matches the 0-4 degree requirement). Finally, in the front-end control, only the original erroneous dropdown box is replaced with carrier B, while retaining the user's other compliant options.

[0073] Please refer to the details. Figure 1 This is a flowchart illustrating a logistics end-to-end visual collaboration method in an embodiment of this application.

[0074] S101. Perform semantic parsing on the logistics collaboration requirements input by the user, and extract operation intent tags and key logistics entities.

[0075] Among them, logistics collaboration needs refer to the original instruction information that users input through text, voice, or form summaries describing their logistics business requirements, such as "arrange for a batch of cold chain goods to be shipped from the Shanghai warehouse to the Chengdu warehouse for a certain customer"; operation intent tags refer to semantic identifiers used to characterize the type of logistics business action that users want to perform, such as placing an order, re-delivery, uploading return receipts, and scheduling returns; logistics key entities refer to the core information units related to specific business objects extracted from user collaboration needs, including but not limited to named entities with business attributes such as shipper, consignee, product category, transportation mode, time window, and origin and destination addresses.

[0076] Specifically, after receiving a user's logistics collaboration request, the system inputs the request text into a pre-trained AI big data model. The big data model then extracts the business semantic information contained in the text based on its internal attention mechanism and encoding layer. On one hand, the system guides the big data model to classify the action semantics of the overall statement, summarize the operation intent label corresponding to the current request, and extract the hidden state features of the corresponding operation intent label in the last layer of the big data model's network as the intent representation vector of that intent label. Alternatively, the classified operation intent label is input into a preset word vector embedding matrix to map and obtain the corresponding multi-dimensional intent representation vector, which is used for subsequent matching of the logistics human-computer interaction interface to be rendered and the comprehensive matching degree calculation of subsequent affected fields. On the other hand, the system guides the big data model to identify key phrases expressing specific business objects in the text, and classifies and normalizes these phrases according to preset logistics business entity types to obtain a structured set of key logistics entities.

[0077] In some embodiments, the system can first perform text cleaning and sentence segmentation on the user's input logistics collaboration requirements, remove redundant characters, and transcribe the voice input into text; then, the processed text is concatenated with a preset prompt word template to form guided input and delivered to a pre-trained AI large model, so that the model can output results simultaneously on the intent classification head and the named entity recognition head; finally, the intent classification results output by the large model are mapped and aligned with a preset operation intent label vocabulary to generate an intent representation vector, and the extracted entity phrases are normalized according to the logistics business entity dictionary to obtain the final operation intent labels and logistics key entities.

[0078] S102. Match the logistics human-machine interface to be rendered according to the operation intent label, and construct the corresponding candidate information set for each field to be filled in the logistics human-machine interface according to the key logistics entities and the pre-acquired logistics historical operation data.

[0079] Among them, the logistics human-computer interaction interface refers to the visual front-end page that the system displays to users to complete specific logistics business operations. Each operation intent label corresponds to one or a set of specific layout interaction interface templates; the fields to be filled refer to the input controls in the logistics human-computer interaction interface that require users or the system to fill in specific business values, such as drop-down boxes, text boxes, date pickers, etc.; logistics historical operation data refers to the structured historical records of logistics orders, node configurations, parameter fillings, etc. generated by the same or similar users in past business processes that are pre-stored in the system; the candidate information set refers to the set of candidate values ​​that the system prepares in advance for each field to be filled, based on key logistics entities and logistics historical operation data, which can be selected by users or used by the system for preliminary filling.

[0080] Specifically, after obtaining the operation intent tag, the system retrieves a matching logistics human-computer interaction interface template from a pre-set interface template library, determining its corresponding page layout, field composition, and field attribute constraints. After determining the interface to be rendered, the system performs a correlation search on historical logistics operation data for each field to be filled, using key logistics entities as search criteria. It then filters historical values ​​relevant to the current business based on field type and semantics. Further, the system performs screening and distance sampling on the retrieved historical values ​​based on early warning trends and historical data feature matrices to limit the number of candidates and establish representativeness. Finally, the system constructs an independent set of candidate information for each field to be filled in the interface, serving as the basis for subsequent preliminary filling calculations and user-selected input.

[0081] S103. Based on the pre-stored logistics data, the topology depends on the candidate information. Each candidate information in the candidate information set is evaluated and calculated to determine the first target candidate information that has the least impact on the change of other fields to be filled.

[0082] Among them, logistics data dependency topology refers to a directed graph structure that is pre-built and stored in the system to depict the causal constraints and linkages between various business fields. The nodes in the graph represent the fields to be filled, and the edges represent the business constraint transmission relationships between the nodes. Candidate information refers to the specific selectable value entries in the candidate information set. The impact of the change range refers to the degree to which the available candidate range of other fields to be filled is compressed when a certain field is assigned a certain candidate information and is transmitted downstream according to the data dependency topology.

[0083] Specifically, for each field to be filled, the system injects each candidate information from its candidate information set as hypothetical input into the logistics data dependency topology, simulating the constraint changes that occur as the candidate information propagates downstream along the topology path. During the simulation, the system evaluates the impact of the candidate information on other fields to be filled based on graph traversal algorithms (such as breadth-first search or depth-first search) and pre-attached business constraint rules in the logistics data dependency topology. This includes, but is not limited to, the degree to which some options in the candidate sets of other fields become invalid, and the degree to which the flexibility of subsequent field values ​​is compressed. The system summarizes and quantifies the overall impact of each candidate information, and performs a horizontal comparison in the candidate information set corresponding to the same field to be filled. Finally, the candidate information that causes the smallest impact is determined as the first target candidate information. This first target candidate information is then used to initially fill and display the corresponding field to be filled in the logistics human-computer interaction interface, presenting the user with an initial form solution that minimizes interference with surrounding fields.

[0084] In some embodiments, the system can first substitute the candidate information from its candidate information set into the corresponding node position in the logistics data dependency topology for each field to be filled; then, starting from that node, it traverses adjacent fields downstream according to the causal relationship of the edges in the topology, and counts the number of downstream options eliminated by the constraints of the candidate information based on the business constraint rules attached to each downstream node; finally, it sorts the downstream elimination quantities obtained by accumulating the candidate information for each candidate information, selects the candidate information with the smallest accumulated value as the first target candidate information, and performs the initial interface filling. It is understood that other methods can also be used to evaluate and determine the candidate information with the least impact on the changes of other fields to be filled, such as by introducing field importance weights to perform weighted aggregation of the impact of each downstream field, or by using graph neural networks to predict the transmission impact of candidate information, which is not limited here.

[0085] The logistics data dependency topology in this embodiment can be generated by the system through structured transformation based on the field cascading verification rules configured in the logistics business rule base, or it can be pre-modeled by business experts based on actual business constraints and stored in the system's topology database. This topology serves as the basic data structure for the operation of this solution and is pre-built independently of the execution process of this method, and is not limited here.

[0086] S104. In response to the user's modification operation on the target field to be filled through the logistics human-computer interaction interface, the derived constraint boundary generated by the new content to be filled corresponding to the modification operation is extracted based on the logistics data dependency topology, and the affected field that conflicts with the derived constraint boundary is located.

[0087] Among them, the modification operation refers to the user's interface interaction behavior such as changing, overwriting, or reselecting a specific field based on the initial filling result on the interface; the target field to be filled refers to the field directly affected by this modification operation, that is, the field object actively adjusted by the user; the derived constraint boundary refers to the set of multi-dimensional attribute constraints formed by the new filling content being deduced downstream along the logistics data dependency topology, which is used to limit the range of legal values ​​of downstream fields; the affected field refers to the downstream field that receives the derived constraint boundary through the logistics data dependency topology due to the modification of the target field to be filled and is detected to have at least one attribute dimension conflicting with the rule.

[0088] This step is executed after the system has completed the initial data filling and presented the interface to the user, when the user actively modifies a field. Specifically, after the system's front-end event listener detects that the user has modified the target field to be filled, it retrieves the initial filled content and the new filled content before modification from the cache. The node corresponding to the target field in the logistics data dependency topology is used as the starting point for this inference. The system further traverses downstream along the causal pointing edges starting from this node in the topology. Combining the business constraint rules pre-attached to downstream nodes, it performs condition tree or logical table-driven inference calculations on the attribute bias caused by the new filled content compared to the initial filled content, thereby generating a derived constraint boundary composed of multi-dimensional attribute constraints. This derived constraint boundary is used to characterize the range of legal candidate data allowed for each field downstream after the impact of this modification operation.

[0089] After generating the derived constraint boundary, the system will perform rule-by-rule verification on the current preliminary filling content of the downstream fields covered by it. Fields with at least one attribute dimension that fails to meet the derived constraint boundary will be identified and marked as affected fields, and will be used as processing objects for subsequent local replacement and attribute extraction.

[0090] Optionally, the system can first perform a difference comparison between the initial and new content of the target field to be filled, extracting the logistics attribute bias that has changed between the two; then, using the topology node corresponding to the target field to be filled as a benchmark, it traverses downstream nodes along the causal pointing edges in the logistics data dependency topology, and loads the business constraint rules attached to the downstream nodes; finally, it substitutes the logistics attribute bias into the business constraint rules of each downstream node to extrapolate the impact range, forming a derived constraint boundary composed of multi-dimensional attribute constraints; then, it compares the current initial content of the downstream fields one by one with the derived constraint boundary, locating fields with dimensional conflicts as affected fields. It is understood that other methods can also be used to generate the derived constraint boundary and locate the affected fields, which is not limited here.

[0091] It should be noted that, in this embodiment, the affected field refers to a downstream field whose initial content has at least one dimension that conflicts with the derived constraint boundary in the multidimensional attribute system; the values ​​of other attribute dimensions in the affected field, excluding the conflicting dimension, may still remain valid. This step is based on this characteristic, and precisely extracts the valid attribute dimension values ​​that can be retained from the affected field.

[0092] S105. Based on the attribute dimension values ​​in the initial filling content of the affected fields that do not conflict with the derived constraint boundaries, construct the retained attribute features.

[0093] Among them, attribute dimension refers to multiple independent feature axes describing the business object corresponding to the affected field. For example, for the field of transportation plan, its attribute dimension may include carrier, transportation mode, timeliness level, vehicle model specification, temperature control requirements, etc.; attribute dimension value refers to the specific business value corresponding to a specific attribute dimension in the initial content of the affected field; retained attribute features refer to the feature set composed of all attribute dimensions and their values ​​in the initial content of the affected field that do not conflict with the rules of the derived constraint boundary.

[0094] Specifically, for each identified affected field, the system examines the values ​​of its initial populated content across each attribute dimension (e.g., performing key-value pair verification for multidimensional data structures). For each attribute dimension, the system determines whether the current value is still within the valid range based on the derived constraint boundaries generated in step S104: if the value of an attribute dimension does not conflict with the corresponding constraint conditions in the derived constraint boundaries, the system marks the attribute dimension and its value as a valid dimension that can be retained; if the value of an attribute dimension exceeds the valid range defined by the derived constraint boundaries, the system marks the attribute dimension as a conflicting dimension that needs to be reselected. After determining all dimensions, the system encapsulates all attribute dimensions marked as retainable and their corresponding values ​​into vectors or hash arrays, and then integrates them into the retained attribute features of the affected field.

[0095] This step ensures that downstream fields ultimately meet the new constraints while extending most of the previously approved sub-attributes to the replacement results, thus avoiding the drawback of discarding all configured values ​​in a "one-size-fits-all" manner under the traditional cascading clear mechanism.

[0096] To facilitate understanding, let's take an order form modification scenario in a logistics system as an example: Suppose the user's original "Transportation Method (upstream modification field)" was "Full Truckload Refrigerated," which determines that the system initially filled in "A Cold Chain Logistics" for the subordinate derivative field "Carrier," and "Temperature Control Requirement" was initially filled in as "0 to 4 degrees Celsius." When the user changes the "Transportation Method" to "Less-than-Truckload (LTL) Cargo," the system deduces along the topology and finds that "Carrier: A Cold Chain Logistics" only supports full truckload business, which conflicts with the new transportation method; however, "Temperature Control Requirement: 0 to 4 degrees Celsius" is still legal and compliant in LTL cargo (i.e., no rule conflict occurs). At this point, the system will not directly clear all downstream form parameters, but will extract "temperature control requirement: 0 to 4 degrees Celsius" as a "retained attribute feature". When the system reselects a carrier from the candidate set for "less-than-truckload (LTL) carpooling", it will use the retained attribute "0 to 4 degrees Celsius" to intelligently find a new carrier that also handles LTL and can meet the temperature control requirement (second target candidate information). The temperature control field is smoothly retained throughout the process and does not require the user to re-enter it.

[0097] S106. Select candidate information that satisfies the derived constraint boundary from the candidate information set corresponding to the affected field. Calculate the feature distance of the selected candidate information based on the retained attribute features and intent representation vector. Select the candidate information with the smallest feature distance as the second target candidate information with the highest comprehensive matching degree to drive the rendering node corresponding to the affected field to perform local data update and replacement display.

[0098] The second target candidate information refers to the final candidate value obtained by re-filtering the affected field from its original candidate information set, which is used to replace the original initial content of the field; the replacement display refers to the process by which the system replaces the original initial content of the affected field with the second target candidate information on the front-end interface and re-renders it to the user.

[0099] Specifically, for each affected field, the system first filters out candidate information from its original candidate information set that conflicts with any attribute dimension of the derived constraint boundary, retaining only candidates that are valid in all dimensions as the available options. Based on this, the system evaluates the matching degree of each valid candidate information with the retained attribute features of the affected field and the current operation intent label: the system extracts and converts the retained attribute features into vector representations, and fuses them with the current operation intent representation vector. Then, it calculates the feature distance (such as Euclidean distance or cosine distance) between the business state vector corresponding to the selected valid candidate information and the fused vector, quantifying the comprehensive matching degree based on the magnitude of this feature distance. The system ultimately selects the candidate information with the smallest feature distance as the second target candidate information with the highest comprehensive matching degree. This ensures that the automatically replaced option retains the user's original preferences while also best matching the modified new intent. This option then replaces the original initial content of the affected field and is re-rendered to the logistics human-machine interface, ensuring compliance while preserving the user's original preferences as much as possible.

[0100] Optionally, the system can first remove any candidates that conflict with the derived constraint boundary from the candidate information set corresponding to the affected field to obtain a verification candidate set; then substitute the candidate information in the verification candidate set into the node position of the affected field in the logistics data dependency topology and calculate the comprehensive matching degree by combining the retained attribute features and operation intention labels; finally, select the candidate information with the highest comprehensive matching degree as the second target candidate information, and replace the displayed content of the affected field with this information on the front-end interface.

[0101] Specifically, to achieve distance calculation for multimodal data, the system can remove invalid items that violate derived constraint boundaries from the candidate information set of the affected field, forming a valid candidate subset. Subsequently, the system calls a pre-built multimodal coding network (such as a multilayer perceptron or graph embedding model) to map the structured preserved attribute features and the intent representation vector of the text domain into a continuous vector space of the same dimension, fusing them to generate a baseline reference vector; simultaneously, each candidate information in the valid candidate subset is also encoded as a candidate vector. Next, the system uses cosine similarity or Euclidean distance algorithms to calculate the relative spatial distance, i.e., the feature distance, between the baseline reference vector and each candidate vector.

[0102] Next, the system enumerates the candidate information in ascending order of feature distance, selecting the first candidate with the smallest distance value as the second target candidate. Subsequently, the system backend sends an asynchronous update command to the frontend. The frontend accurately locates the DOM (Document Object Model) rendering node corresponding to the affected field based on the component identifier carried in the command, and performs a non-refreshing partial view redraw and replacement display, thus seamlessly completing the system's replacement and rescue.

[0103] To better understand the above derivation and partial replacement process of retained attribute features, a specific scenario of multimodal transport waybill information entry is used as an example: Assume the target field to be filled (upstream backbone field) is "Transportation Mode," and the system pre-fills it with "Full Truckload Refrigerated." The causally related affected fields (downstream linked fields) include "Carrier" (initially filled as 'A Cold Chain Line') and "Temperature Control Requirements" (initially filled as '0 to 4 degrees Celsius'). When the user changes the "Transportation Mode" to "Less-than-Truckload (LTL)" due to business changes, the system derives new constraints accordingly. Based on the derived constraint boundary, "A Cold Chain Line" becomes invalid due to a rule conflict because it only handles full truckload business; however, "0 to 4 degrees Celsius" remains valid in LTL LTL business (i.e., no conflict occurs). Therefore, the system does not use the traditional full form clearing mechanism but extracts "0 to 4 degrees Celsius" as a retained attribute feature. When re-screening carriers, the system, based on this retention attribute, automatically matches a new carrier B from the candidate pool that supports less-than-truckload (LTL) freight, has a similar distance requirement, and best matches the 0-4 degree Celsius temperature control requirement, and silently replaces it with the second target candidate. In this way, the user-confirmed temperature control rules are successfully retained, eliminating the tedious manual re-entry process.

[0104] It should be noted that the replacement display performed in this step is the final display state of the affected fields in application scenarios that do not trigger the subsequent visual verification process; however, in application scenarios that trigger the visual verification process (as described in steps S107 to S110 below), the replacement display in this step can be used as a pre-presentation state pending user confirmation, and will finally take effect through the update confirmation interaction in S110.

[0105] S107. Extract the feature deviation between the initial filling content corresponding to the affected field and the second target candidate information, and perform weighted calculation with the feature deviation and the preset business weight matrix to obtain the verification priority parameter.

[0106] Among them, feature deviation refers to the degree of difference between the two contents before and after the affected field is replaced in each attribute dimension; business weight matrix refers to the weight structure composed of the business importance coefficients of each node in the logistics data dependency topology, which is used to reflect the criticality of different fields in the entire business chain; business importance coefficient refers to the numerical parameter preset for each field, which represents the magnitude of the impact of the field on business compliance and execution results; verification priority parameter refers to the priority quantitative indicator obtained by weighting the feature deviation and business weight matrix, which is used to measure whether the user needs to be prompted to actively verify in this replacement.

[0107] Specifically, for each affected field that has been replaced, the system extracts the differences in values ​​across various attribute dimensions between the original initial fill content and the second target candidate information. These differences are then aggregated and quantified into feature deviations according to a preset method, such as calculating the feature deviation between the original initial fill content and the second target candidate information based on character editing distance (e.g., Levenshtein distance) or semantic vector difference. Based on this, the system loads a pre-constructed business weight matrix. Each item in this matrix corresponds to the business importance coefficient of a field node in the logistics data dependency topology. The coefficients for key fields (e.g., timeliness level, temperature control range, compliance approval, etc.) are typically significantly higher than those for general attribute fields. For example, core financial / performance fields such as billing mode and carrier name correspond to high weights, while remark fields correspond to low weights. The system then performs a weighted calculation by adding the feature deviation to the weighted items corresponding to the affected field in the business weight matrix, thereby obtaining the verification priority parameter for that affected field.

[0108] Optionally, the system can first vectorize the initial content of the affected field and the second target candidate information into attribute vectors of the same dimension; then calculate the difference between the two attribute vectors by dimension and aggregate them into the feature deviation value corresponding to the affected field according to a preset method; finally, extract the corresponding business importance coefficient from the business weight matrix according to the position of the affected field in the logistics data dependency topology, and multiply it with the feature deviation to obtain the verification priority parameter.

[0109] It should be noted that in this embodiment, the feature deviation is used to measure the degree of difference between the two contents before and after the affected field is replaced. This is different from the semantic discrete bias used in the aforementioned steps to measure the degree of semantic drift before and after the target field to be filled is modified. The former is used to determine whether to trigger visual verification, while the latter is used to determine the generation method of the baseline state vector.

[0110] S108. When the verification priority parameter is greater than the preset threshold, the directed transmission path from the target field to be filled to the affected field in the logistics data dependency topology is mapped to the screen coordinate system of the logistics interaction interface to generate a page linkage guidance trajectory.

[0111] Among them, the preset threshold refers to the verification priority threshold pre-configured by the system to determine whether a replacement needs to trigger a visual verification prompt; the directed transmission path refers to the ordered sequence of nodes in the logistics data dependency topology that starts from the node corresponding to the target field to be filled and reaches the node corresponding to the affected field along the causal pointing edge; the screen coordinate system refers to the two-dimensional pixel coordinate system used by the logistics interactive interface when rendering it on the front end; the page linkage guide trajectory refers to the visual guide line formed by connecting the positions of the corresponding fields of each node on the directed transmission path on the interface in sequence, which can be dynamically drawn and presented on the front end.

[0112] Specifically, the system compares the verification priority parameter corresponding to the affected field with a preset threshold. When the parameter is less than or equal to the threshold, it determines that the replacement does not require special prompting to the user and the replacement result can be retained silently. When the parameter is greater than the threshold, it determines that the replacement is a high-priority linkage change that needs to be actively perceived by the user. After determining that a visual prompt needs to be triggered, the system will trace back its directed transmission path in the logistics data dependency topology, starting from the node corresponding to the target field to be filled and ending at the node corresponding to the affected field, and sequentially obtain the rendering position of the field corresponding to all nodes on the path in the logistics interaction interface. Then, the system will sequentially map these rendering positions to specific coordinate points in the screen coordinate system and connect them according to the path order to form a page linkage guidance trajectory. This trajectory can intuitively express the transmission relationship from the user's modification point to the affected field on the front end, providing a positional basis for the subsequent rendering of the floating component.

[0113] S109. Render the floating interface component based on the page linkage guide trajectory, and output the initial fill content and the second target candidate information in the floating interface component to generate a change comparison view.

[0114] Among them, the floating interface component refers to the floating visual component that is superimposed on the regular layer of the logistics interaction interface and dynamically presented according to the page linkage guide trajectory, used to carry explicit prompts of this linkage change; the parallel output refers to displaying the two contents before and after the affected field is replaced side by side in the same floating component, so that users can directly compare them; the change comparison view refers to the comparison display view composed of the initial filled content and the second target candidate information presented side by side, used to support users to confirm the update.

[0115] Specifically, the system relies on a page-linked guidance trajectory, overlaying and rendering floating interface components on the logistics interaction interface along key nodes of the trajectory. The components are generally positioned adjacent to the area of ​​the affected field and are visually associated with the target field to be filled through the guidance trajectory, allowing users to clearly identify the source and end point of this linked change. The system further injects the original preliminary filling content of the affected field and the second target candidate information obtained from the replacement into the floating interface component in a side-by-side or top-bottom juxtaposition format. Visual enhancement markers can be added to the dimensions of difference attributes, thereby forming a complete change comparison view in the floating component.

[0116] Optionally, the system can first create a floating interface component on the front-end layer based on the endpoint of the page linkage guide trajectory, and visually bind the guide trajectory to the component; then fill the initial filling content corresponding to the affected field and the second target candidate information into the floating interface component according to the preset juxtaposition template; finally, apply color or symbol markings to the differences between the two in the attribute dimension, so that a complete change comparison view is formed inside the floating interface component and presented to the user.

[0117] S110. In response to the update confirmation interaction command for the change comparison view within the page linkage guide trajectory range, update and confirm the affected fields.

[0118] Among them, the update confirmation interaction instruction refers to the interactive operation performed by the user on the changed comparison view within the coverage area of ​​the page linkage guide trajectory, which is used to express acceptance of the replacement result, such as clicking the confirmation button, selecting the second target candidate information, or performing a preset gesture on the comparison view; update confirmation refers to the process by which the system formally solidifies the displayed content of the affected field and the underlying data state into the second target candidate information after receiving the above interaction instruction.

[0119] Specifically, after rendering the page linkage guide trajectory and the change comparison view, the system listens for user interaction within the interface area covered by the guide trajectory, filtering out routine operations outside this area and only responding to interaction commands on the change comparison view that express confirmation intent. When the system detects that the user has made an update confirmation interaction command on the change comparison view within this area, it considers that the user has completed the verification and acceptance of this linkage change. The system then officially updates the affected fields as the second target candidate information at the underlying data level, simultaneously releasing or closing the corresponding floating interface components and page linkage guide trajectory, and fixing the final value of the field in a stable display form on the interactive interface; conversely, if the user refuses to update, the system supports triggering in-situ undo and rollback.

[0120] In some embodiments, the system can first bind an event listener to the change comparison view on the front end, which is limited to the coverage area of ​​the page linkage guide trajectory; then, when the system detects that the user performs a confirmation-type interaction operation within this area, it can identify the operation as an update confirmation interaction instruction; finally, it can update the value of the affected field in the underlying data model to the second target candidate information and close the floating interface component and the page linkage guide trajectory.

[0121] Based on the embodiments corresponding to steps S101 to S110, in order to further improve the refinement of each stage of candidate information construction, preliminary filling, and derived constraint deduction, this application provides detailed embodiments corresponding to S201 to S212 as a more specific implementation of the aforementioned steps S102 to S106. After implementing this embodiment, without increasing the user's manual verification burden, the display scale of front-end candidate options can be further reduced, the probability of secondary modifications caused by preliminary filling results can be decreased, and the update results of downstream fields can be made closer to the user's current actual business intent. The method provided in this embodiment will be described in further detail below. Please refer to... Figure 2 This is another flowchart illustrating a logistics end-to-end visual collaboration method in this application embodiment.

[0122] S201. Based on the correlation search of key logistics entities in historical logistics operation data, an initial candidate set is obtained.

[0123] Among them, the related retrieval refers to the process by which the system uses key logistics entities as retrieval anchors, matches and queries along the related attributes of business objects in the historical operation data of logistics, and filters out historical value entries that are relevant to the current business needs in dimensions such as shipper, consignee, product category, transportation mode, and time window.

[0124] This step is performed after the system completes the extraction of intent tags and key logistics entities in step S101, and matches the logistics human-computer interaction interface to be rendered based on the operation intent tags. It is the initial stage of preparing to build a candidate information set for each field to be filled in the interface.

[0125] Specifically, after determining the logistics human-computer interaction interface to be rendered, the system sequentially traverses each field to be filled on the interface and determines the key entity dimensions to be used for this association retrieval based on the semantic type of each field. The system uses entity values ​​in the logistics key entity set that match the semantics of the field as retrieval conditions, and performs cross-matching along data fragments such as historical orders, historical node configurations, and historical parameter fillings in the logistics historical operation data. It extracts value entries that have appeared in contexts similar to the current business in terms of shipper, consignee, product category, or transportation mode. At the same time, the system retains a snapshot of the associated attributes of each candidate value in the historical data, such as the carrier, vehicle type, timeliness level, and temperature control range that the value is associated with in historical business, so that subsequent steps can use these attribute snapshots to perform anomaly removal and differential sampling. After completing the above retrieval, the system groups and collects all the obtained candidate value entries by field to form the initial candidate set corresponding to each field to be filled.

[0126] In some embodiments, the system can first filter out a subset of entities related to the field from the set of key logistics entities based on the semantic type of the field to be filled; then, using the subset of entities as search conditions, perform a multi-field joint association query in the structured historical business table corresponding to the historical logistics operation data to recall all historical value records of at least one dimension in the matched entity subset; finally, the recall results are deduplicated and merged according to the original values, and a snapshot of the association attributes of each value in the historical business is retained to form the initial candidate set corresponding to the field to be filled.

[0127] S202. Based on the real-time status warning data of the current logistics network environment, candidate information with abnormal attribute characteristics in the initial candidate set is eliminated to obtain a usable candidate set.

[0128] The current logistics network environment refers to the operational status of logistics resources during the current business execution period, which is composed of the capacity status of various carriers, warehouse receiving and dispatching capacity, trunk road network traffic conditions, transportation vehicle operation status, and regional control measures. Real-time status early warning data refers to the structured early warning information obtained by the system in real time from relevant operational monitoring channels to describe the existence of abnormal or restricted situations in the current logistics network environment, such as road closures on a trunk line, warehouse overload and suspension of receiving, temporary suspension of operation of a certain vehicle type, and traffic restrictions in a certain area. Abnormal attribute features refer to the feature fields carried by candidate information in its associated attribute snapshot that can establish a matching relationship with the abnormal events in the real-time status early warning data in terms of attribute dimension. The available candidate set refers to the set of candidate values ​​that are truly available in the current logistics network environment after the initial candidate set is removed from all candidate information with abnormal attribute features.

[0129] Specifically, after obtaining the initial candidate set output by S201, the system synchronously retrieves real-time status warning data of the current logistics network environment and performs structured parsing of the warning data according to the warning object type, obtaining a series of warning entries in the form of "object-abnormal attribute-impact scope". Subsequently, the system iterates through each candidate information in the initial candidate set, comparing the snapshot of the associated attributes attached to each candidate information with the parsed warning entries: if the attribute carried by the candidate information in any dimension such as carrier, vehicle type, warehouse, route, or time period matches the object and its abnormal attributes pointed to by a certain warning entry, the system determines that the candidate information is currently unusable and removes it from the initial candidate set; if the snapshot of the associated attributes of the candidate information does not match any of the warning entries, the system determines that the candidate information is not currently affected by abnormal factors in the current logistics network environment and retains it as a usable entry. After completing the comparison of all candidate information, the system aggregates all the retained candidate information to form a usable candidate set for each field to be filled.

[0130] In some embodiments, the system can first call the early warning data interface to retrieve real-time status early warning data of the current logistics network environment, and then structure the early warning data according to object dimensions such as carrier, warehouse, route, and vehicle type; then iterates through each candidate information in the initial candidate set, performs attribute-level matching between the snapshot of its associated attributes and the structured early warning entries, and marks the candidate information that hits an anomaly in any dimension; finally, all candidate information that is not marked as an anomaly is gathered into a usable candidate set.

[0131] S203. Based on the feature distance between the business feature vectors corresponding to each candidate information in the available candidate set, sample the candidate information to generate a candidate information set that meets the preset capacity upper limit threshold.

[0132] Among them, the business feature vector refers to the multi-dimensional vector used to represent the business profile of each candidate information in a unified vector space after extracting numerical or coded indicators along multiple business dimensions of its associated attributes for each candidate information in the available candidate set; the feature distance refers to the distance index that measures the degree of difference between the corresponding business feature vectors of two candidate information in each orthogonal attribute axis in the unified vector space. The larger the value of this index, the more obvious the difference between the two candidate information in the business profile; the capacity upper limit threshold refers to the upper limit parameter of the number of fields set in advance for each type of field to be filled, which is used to constrain the length of the final candidate information set.

[0133] Specifically, for each available candidate set corresponding to a field to be filled, the system first defines the corresponding feature extraction rules based on the business attribute system of the field, and maps each candidate information in the available candidate set to a business feature vector of a unified dimension according to the rules; then, the system calculates the feature distance between candidate information pairwise in the vector space to obtain a distance relationship matrix that characterizes the degree of differentiation between candidate information.

[0134] Based on this, the system loads the pre-configured capacity limit threshold for the current field and simultaneously loads the preset differentiation threshold. In differentiated sampling, the system uses the feature distance between candidate information as an exclusion criterion. Only when the feature distance between a candidate information and all previously selected candidate information is greater than the differentiation threshold is it included in the candidate information set. At the same time, the capacity limit threshold constrains the final number of selected items, thereby generating a candidate information set that satisfies both the requirement that the pairwise feature distance is greater than the differentiation threshold and the capacity limit. By calculating the feature distance to ensure the degree of difference between samples, highly homogeneous (i.e., semantically close) redundant business items in the drop-down list can be filtered out, thereby reducing the number of times users have to search through long lists on the front end, while ensuring that the candidate information set maintains sufficient representativeness in terms of business coverage as much as possible.

[0135] In some embodiments, the system can first extract numerical or encoded attribute indicators for each candidate information in the available candidate set according to the business attribute system of the field to be filled, and then concatenate them into a business feature vector of a unified dimension; then calculate the feature distance between candidate information pairwise in the vector space to obtain a distance relationship matrix; finally, with the capacity upper limit threshold as a constraint, adopt a greedy or near-optimal sampling strategy to iteratively select a set of candidate information in the available candidate set, so that the sum of the corresponding values ​​of the selected candidate information in the distance relationship matrix reaches the maximum value in the sampling combination of the same capacity, and organize the set of candidate information into the final candidate information set.

[0136] S204. Input each candidate information in the candidate information set as a trigger node into the logistics data dependency topology to simulate constraint transmission and extract the cascading influence path that affects other fields to be filled.

[0137] In this context, the trigger node refers to the topological node that is injected into the logistics data dependency topology as a signal source during a constraint propagation simulation, and is used to release the constraint influence to its downstream nodes step by step. In the current simulation, this node corresponds to the field to be filled with candidate information hypothetically. The constraint propagation simulation refers to the simulation calculation process in which the system starts from the trigger node in the logistics data dependency topology, and releases the constraints carried by the candidate information to the downstream nodes step by step along the causal pointing edges, and continuously deduces the changes in the legal values ​​of the downstream fields based on the business constraint rules attached to the downstream nodes. The cascading influence path refers to the directed node sequence that starts from the trigger node, passes through several causal pointing edges, and finally covers the nodes corresponding to other fields to be filled in a constraint propagation simulation.

[0138] Specifically, when performing a preliminary filling assessment on a field to be filled, the system retrieves each candidate information item from the candidate information set corresponding to that field and inputs it as the trigger node into the logistics data dependency topology for this simulation. During the simulation, the system treats the attribute values ​​carried by the candidate information as already filled into the field corresponding to the trigger node, and releases constraint signals level by level to downstream nodes along the causal pointing edges starting from that node in the logistics data dependency topology. Each time a causal pointing edge is passed, the system reads the business constraint rules attached to the target node and, based on the constraints currently passed to that node, infers whether the legal value range of the field corresponding to that node has narrowed, and whether that node will continue to pass new derived constraints to further downstream nodes. The entire simulation process unfolds hierarchically along the logistics data dependency topology until the transmitted signals attenuate to the point where they can no longer reach new fields to be filled. After the simulation ends, the system backtracks along the connection lines of the nodes that actually caused the constraint changes, extracting all directed node sequences starting from the trigger node and ultimately covering other fields to be filled, as the cascading influence path corresponding to that candidate information.

[0139] In some embodiments, the system can first extract a candidate information from the candidate information set and attach it to the node position of the corresponding field in the logistics data dependency topology; then, using the node as the root, it can traverse downstream along the causal pointing edge based on the breadth-first or depth-first strategy, and read the business constraint rules attached to the target node each time it crosses the causal pointing edge, and record the narrowing of the legal value range caused by the constraint on the node; finally, it can concatenate all the downstream field nodes that have generated value range changes into one or more directed node sequences, and summarize them into the cascading influence path corresponding to the candidate information.

[0140] S205. For each candidate information in the candidate information set, based on the cascading influence path, count the number of options in the candidate information set that fail due to constraint conflicts in the corresponding candidate information set for other fields to be filled affected by the candidate information, and then determine the first target candidate information with the minimum space compression cost.

[0141] Among them, the number of options refers to the number of option entries in the candidate information set corresponding to other fields to be filled that are affected by a constraint transmission simulation, which cannot continue to be legally selected due to the derivative constraints caused by the current candidate information; the space compression cost refers to the cost parameter used to characterize the degree to which the candidate information occupies the space of legal options in the surrounding fields when it is used as the initial filling result, obtained by summarizing the number of invalid options corresponding to all other fields to be filled affected by a candidate information in the simulation according to a preset method; the first target candidate information refers to the final candidate information with the minimum space compression cost in the candidate information set, which is used to perform the initial filling and display of the corresponding fields to be filled in the logistics human-computer interaction interface.

[0142] Specifically, for each candidate information in the candidate information set corresponding to the current field to be filled, the system first extracts the cascading influence path calculated in S204, and sequentially locates other affected fields to be filled along this path. For each affected field, the system applies the derived constraints caused by the candidate information to the candidate information set corresponding to that field, and compares the attribute value of each candidate entry in the set with the derived constraints to see if there is a dimensional conflict. Entries that cannot satisfy the constraint are counted as failed options, and the total number of failed options corresponding to the affected field is finally summarized. The system then accumulates and aggregates the number of failed options corresponding to all other fields affected by the current candidate information according to a preset method to obtain the space compression cost corresponding to the current candidate information.

[0143] After the system completes the spatial compression cost calculation for all candidate information in the candidate information set, it compares the cost values ​​horizontally within the same candidate information set, selects the candidate information with the lowest cost as the first target candidate information, and pushes the first target candidate information to the front end for preliminary filling and display of the corresponding fields to be filled in the logistics human-machine interaction interface.

[0144] S206. Compare the new content corresponding to the modification operation with the original preliminary content of the target field to be filled, and extract the logistics attribute bias amount that has changed.

[0145] Among them, differential comparison refers to the process by which the system compares the two values ​​of the target field to be filled before and after being modified by the user across multiple attribute dimensions, identifies the dimensions that have changed, and quantifies the magnitude of the change. The logistics attribute bias refers to the structured result extracted through differential comparison, which is used to characterize the direction and magnitude of the attribute value change caused by the modification operation. It can include the set of changed dimensions and the value deviation information of the transition from the initial filled content to the new filled content in each dimension.

[0146] Specifically, after the system detects a user's modification operation on a target field to be filled, it simultaneously obtains the initial content before the modification and the newly submitted content. Both are then mapped to the corresponding attribute dimension system, making them comparable within the same attribute coordinate system. Subsequently, the system compares the values ​​of the initial and new content along each attribute dimension: for dimensions where the two contents have the same value, the system marks them as stable dimensions not affected by the modification and excludes them from the logistics attribute bias calculation; for dimensions where the two contents have different values, the system further measures the direction and magnitude of the transition from the initial to the new content, such as narrowing the temperature control range from "0 to 4 degrees Celsius" to "below -18 degrees Celsius" or changing the transportation method from "full truckload refrigerated" to "less-than-truckload (LTL) carpooling." The system then organizes the dimension name, the value before the change, the value after the change, and the quantified magnitude of the change into a unified bias term for that dimension. The system aggregates all identified biases into logistics attribute biases, which serve as inputs for subsequent steps S207 when traversing downstream nodes in the logistics data dependency topology and S208 when deriving derived constraint boundaries.

[0147] In some embodiments, the system can first retrieve the content of the target field to be filled before and after the modification event is triggered, and then parse the two contents into attribute key-value pair sets organized by dimension according to the attribute dimension system corresponding to the field; then compare the two key-value pair sets one by one along the attribute dimension, skip the dimensions with the same value, and record the dimension name and the specific value before and after the change for the dimensions with different values; finally, encapsulate all the changed dimensions together with the change range into a structured logistics attribute bias object.

[0148] S207. Traverse the downstream nodes in the logistics data dependency topology that have a causal relationship with the target field to be filled, and obtain the corresponding business constraint rules.

[0149] Specifically, after outputting the logistics attribute bias in S206, the system uses the node of the target field to be filled in the logistics data dependency topology as the base node and loads all causal pointing edges in the topology originating from the base node. The system then sequentially visits the adjacent nodes pointed to by these causal pointing edges, registering them one by one as downstream nodes for this simulation. While registering each downstream node, the system reads the pre-configured business constraint rules on that node from the attribute mounting area of ​​the topology node, obtaining structured information such as the constraint attribute dimension, constraint conditions, and legal value shrinkage method declared by the rules. The system then compiles all registered downstream nodes and their mounted business constraint rules into a list to be simulated according to the topology traversal order.

[0150] In some embodiments, the system can first locate the corresponding node in the logistics data dependency topology based on the identifier of the target field to be filled and mark it as the base node; then, starting from the base node, visit the adjacent nodes pointed to by each causal pointing edge in the topology with the node as the exit point, and add them to the downstream node list in turn; finally, read the business constraint rules configured for each downstream node from the attribute mounting area of ​​the topology node, and pair the node identifier with the corresponding rule to form a list to be deduced and deliver it to the next stage.

[0151] It should be noted that the reference node refers to the topological node used as the starting point for traversal during this downstream node traversal process. In this step, it specifically corresponds to the node mapped to the target field to be filled in the logistics data dependency topology. The causal pointing relationship refers to the directed association relationship in the logistics data dependency topology from one node to another, used to express that the change of the upstream field value will have a constraint effect on the legal range of the downstream field. The downstream node refers to the adjacent node in the logistics data dependency topology that has a causal pointing edge with the reference node as the starting point, and will therefore bear the constraint transmitted by the change of the reference node value. The business constraint rule refers to the rule-based expression pre-attached to the downstream node, used to describe how the legal value range of the corresponding field of the node should shrink after receiving the upstream constraint signal, such as the constraint condition with business semantics, such as "when the transportation mode is LTL carpooling, the temperature control range shall not be stricter than -18 degrees Celsius".

[0152] S208. Substitute the logistics attribute bias into the business constraint rules to perform influence interval deduction, generate the derived constraint boundary composed of multi-dimensional attribute constraints, and extract the candidate information that does not conflict with the derived constraint boundary from the candidate information set corresponding to the affected field to obtain the verification candidate set.

[0153] Among them, the influence interval deduction refers to the process by which the system takes the logistics attribute bias as input and substitutes it into the business constraint rules attached to the downstream node, and then calculates the legal value shrinkage interval caused by each of the rule's condition branches and action intervals. The multidimensional attribute constraint conditions refer to a series of legal value restriction conditions output by the influence interval deduction, which correspond to different attribute dimensions of the affected field. Each dimension constraint condition corresponds to one attribute dimension of the affected field. The verification candidate set refers to the set of legal candidate values ​​formed after filtering out candidates that conflict with any dimension of the derived constraint boundary from the candidate information set corresponding to the affected field. This set is used for subsequent benchmark state vector comparison and final screening.

[0154] Specifically, after obtaining the list of items to be simulated output by S207, the system extracts the business constraint rules attached to each downstream node in the list and maps the logistics attribute biases output by S206 into the input parameters of the rules one by one according to the attribute dimensions declared by the rules. Within each business constraint rule, based on the conditional branches to which the bias value falls, the system simulates the contraction of the rule's legal value range for the downstream node in the corresponding attribute dimension, generating attribute constraints for the corresponding dimension. The system then aggregates the attribute constraints derived from multiple rules for the same downstream node into a multi-dimensional attribute constraint set, serving as the derived constraint boundary for the affected field corresponding to that downstream node. The system then performs a line-by-line verification of the candidate information set for the affected field based on the derived constraint boundary, comparing the compatibility of the values ​​of each candidate information with the constraints in each dimension of the derived constraint boundary across all attribute dimensions. Candidates with conflicts in any dimension are eliminated, and candidates that pass all dimension verifications are retained and aggregated into a verification candidate set corresponding to the affected field.

[0155] In some embodiments, the system can first inject the logistics attribute bias into the business constraint rules attached to each downstream node in the list to be simulated according to the rule input parameter specification, and deduce the legal value shrinkage range on each attribute dimension according to the conditional branches of the rules; then, organize the legal value shrinkage ranges derived from each attribute dimension on the same node into a multi-dimensional attribute constraint condition set, which serves as the derived constraint boundary of the affected field corresponding to that node; finally, traverse each candidate in the candidate information set corresponding to the affected field, perform compatibility verification along all attribute dimensions and derived constraint boundaries, and retain and organize the candidates that pass all dimension verifications into a verification candidate set.

[0156] S209. Extract the deviation direction between the new filling content and the initial filling content corresponding to the modification operation in the feature semantic space, and generate the intention drift vector by combining the logistics attribute bias.

[0157] Among them, the feature semantic space refers to the high-dimensional vector space pre-constructed by the system to represent the semantic differences of logistics business field values ​​in the form of continuous vectors. In this space, values ​​with similar business semantics are close to each other, while values ​​with large differences in business semantics are far apart. The intent drift vector refers to the vectorized expression formed by the deviation direction, which represents the direction in which the user's modification has drifted relative to the original initial filling result at the business intent level.

[0158] Specifically, after the system detects a user's modification operation on the target field to be filled and obtains both the initial and new values, it acquires the logistics attribute bias derived from the aforementioned difference comparison. The system then maps both values ​​to a pre-constructed feature semantic space, resulting in two comparable semantic representation vectors within the same space. Further, using the semantic representation vector corresponding to the initial filled content as the starting point and the semantic representation vector corresponding to the new filled content as the ending point, the system calculates the directed difference between the two, extracts the deviation direction from the starting point to the ending point, and combines this deviation direction with the corresponding macro-logistic attribute bias to express it in vectorized form as an intent drift vector.

[0159] In some embodiments, the system can first send the initial filling content and the new filling content corresponding to the target field to be filled into a pre-trained feature semantic encoding module to obtain two semantic representation vectors in a unified feature semantic space; then, taking the vector corresponding to the initial filling content as the starting point and the vector corresponding to the new filling content as the ending point, calculate the dimension-wise difference between the two along each semantic dimension to construct a directed displacement vector from the starting point to the ending point; finally, output the directed displacement vector as the intention drift vector.

[0160] S210. Jointly map the operation intent label and the business attribute dimension of the business constraint rule to the same source representation space to generate the initial intent reference vector, and calculate the feature distance between the intent drift vector and the initial intent reference vector to obtain the semantic discrete bias.

[0161] Among them, the business attribute dimension refers to the attribute dimension declared in the business constraint rules, which is used to constrain the legal values ​​of downstream nodes; the homologous representation space refers to the unified vector space specially constructed by the system that can simultaneously accommodate the semantic representation of the operation intention label and the feature representation of the attribute indicators of the business constraint rules, and make the two directly comparable in this space; the initial intention reference vector refers to the vectorized expression generated by jointly mapping the operation intention label and the attribute indicators of the business constraint rules to the homologous representation space, which is used to characterize the user's current collaboration intention and the original reference state of the downstream constraint rules in the unified space; the semantic discrete bias refers to the feature distance parameter between the intention drift vector and the initial intention reference vector in the homologous representation space, which is used to quantify the degree of deviation of the semantic drift caused by this modification from the original reference intention.

[0162] Specifically, after obtaining the intent drift vector output by S209, the system loads the operation intent label corresponding to this business and the business attribute dimensions from the downstream node business constraint rules already obtained in S207. The semantic encoding of the operation intent label and the business attribute dimensions of each business constraint rule are projected together onto a common-source representation space using a preset mapping function, unifying the two originally heterogeneous semantic information within this space. The system then aggregates the above mapping results along each dimension of the common-source representation space into a single vector, serving as the initial intent reference vector for this operation, characterizing the coupling relationship between the user's initial collaborative intent expressed through the operation intent label and the downstream business constraint rules. The system further calculates the feature distance between the intent drift vector and the initial intent reference vector in the common-source representation space, and outputs this feature distance as a semantic discrete bias, used to quantify whether the modification has deviated semantically from the coupling reference state between the original collaborative intent and the downstream business constraint rules.

[0163] S211. When the semantic discrete bias exceeds the preset redirection threshold, the association indicators of the retained attribute features are constructed into an initial attribute matrix, and the values ​​of each dimension in the initial attribute matrix are weighted according to the weight of the adjacent association edge corresponding to the affected field to generate a local feature weighted matrix.

[0164] Among them, the redirection threshold refers to the critical value set by the system in advance for semantic discrete bias, used to determine whether the semantic drift caused by this modification is large enough to require weighted alignment of the retained attribute features; the initial attribute matrix refers to the matrix representation formed by organizing the values ​​of each non-conflicting attribute dimension contained in the retained attribute features according to a preset arrangement rule, with each row or column in the matrix corresponding to one attribute dimension value in the retained attribute features; the adjacency association edge weight refers to the weight coefficient attached to each causal pointing edge in the logistics data dependency topology with the node corresponding to the affected field as the endpoint, used to characterize the tightness of the business association between the affected field and its adjacent nodes.

[0165] Specifically, after obtaining the semantic discrete bias output by S210, the system compares it with a preset redirection threshold: when the semantic discrete bias is not detected to exceed the preset redirection threshold, the system determines that the semantic drift caused by this modification is relatively small and no core logic jump has occurred. At this time, the system directly generates and outputs the baseline state vector required by the subsequent S212 based on the original retained attribute features; conversely, when the semantic discrete bias is detected to exceed the preset redirection threshold, the system determines that this modification has caused a significant drift at the semantic level and that weighted alignment of the retained attribute features is required.

[0166] After entering the weighted alignment branch, the system numerically maps the non-conflicting attribute dimension values ​​contained in the retained attribute features and organizes them into an initial attribute matrix according to a preset arrangement rule, so that each attribute dimension value in the retained attribute features has a clear position in the matrix. Subsequently, in the logistics data dependency topology, the system uses the node corresponding to the affected field as the query anchor point, extracts the weights of all adjacent related edges attached to the causal pointing edges with that node as the endpoint, and performs alignment mapping between the attribute dimension corresponding to each adjacent related edge weight and the dimension position in the initial attribute matrix. The system then uses the aligned adjacent related edge weights to perform item-by-item weighted allocation of each dimension value in the initial attribute matrix, so that the originally equally parallel dimension values ​​are readjusted into differentiated weighted values ​​in the matrix according to their correlation with the affected field. Finally, it outputs a local feature weighted matrix that is relatively aligned with the local correlation strength of the topology, as the basis for the subsequent generation of the baseline state vector in S212.

[0167] It should be noted that the semantic discrete bias judgment mechanism is introduced to avoid unnecessary realignment operations on the retained attribute features when the user's modification is small (such as a synonymous switch), thereby saving computing resources; while when the user's modification is large (such as a change in the core business mode), the baseline state vector is made closer to the real intention after the modification by weighted alignment, so as to avoid the local replacement result deviating from the user's current intention.

[0168] S212. Based on the node-level characteristics of the logistics data dependency topology, the local feature weighting matrix is ​​transformed into a baseline state vector.

[0169] Among them, node-level features refer to the structured attributes used to characterize the position and weight of nodes in the overall topology, such as the hierarchical relationship of each node in the logistics data dependency topology, the upstream and downstream causal distance, and the business importance of each node; the benchmark state vector refers to the unified vectorized expression formed by weighting the realigned feature matrix according to the node-level features and multi-dimensional vectorization encoding, which is used to compare the Euclidean distance with each candidate state vector in the subsequent candidate information screening stage. This benchmark state vector serves as a comparison benchmark for measuring the fit between each candidate information in the verification candidate set and the current real demand of the affected fields.

[0170] Specifically, after obtaining the local feature weighting matrix output by S211, the system loads the node-level features of the nodes corresponding to the affected fields in the logistics data dependency topology, including the node's hierarchical depth, causal distance to the baseline node, and the node's own business importance coefficient, among other structured information. Based on these hierarchical features, the system determines the positional arrangement and weight allocation of each matrix parameter in the local feature weighting matrix during the vectorization encoding process, ensuring that parameters at key levels or carrying key business semantics occupy corresponding influential positions in the final vector. The system further performs multi-dimensional vectorization encoding on the matrix parameters in the local feature weighting matrix according to the aforementioned positional arrangement and weight allocation, compressing the matrix parameters into a unified-dimensional vectorized expression, i.e., the baseline state vector. In the subsequent screening stage, this baseline state vector is compared with the candidate state vectors extracted from each candidate information in the verification candidate set output by S208 after simulation at the node positions in the logistics data dependency topology using Euclidean distance. Finally, the candidate information corresponding to the smallest Euclidean distance is selected as the second target candidate information, and the affected fields are replaced and displayed.

[0171] The logistics end-to-end visualization and collaborative system of this invention is applied to electronic devices. Figure 3 A schematic diagram of the architecture of an electronic device suitable for implementing embodiments of the present invention is shown.

[0172] It should be noted that, Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0173] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by instructions (computer programs), or by instructions (computer programs) controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. The electronic device of this embodiment includes a storage medium and a processor, wherein the storage medium stores multiple instructions that can be loaded by the processor to execute any step of the method provided in the embodiments of the present invention.

[0174] Specifically, the storage medium and the processor are electrically connected directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected to each other via one or more signal lines. The storage medium stores computer-executable instructions that implement data access control methods, including at least one software functional module that can be stored in the storage medium in the form of software or firmware. The processor executes various functional applications and data processing by running the software program and module stored in the storage medium. The storage medium can be, but is not limited to, Random Access Memory (RAM), Read-Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The storage medium stores the program, and the processor executes the program after receiving the execution instructions.

[0175] Furthermore, the software programs and modules within the aforementioned storage medium may also include an operating system, which may include various software components and / or drivers for managing system tasks (e.g., memory management, storage device control, power management, etc.) and can communicate with various hardware or software components to provide an operating environment for other software components. The processor may be an integrated circuit chip with signal processing capabilities. The aforementioned processor may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc., which can implement or execute the methods, steps, and logic flowcharts disclosed in this embodiment. The general-purpose processor may be a microprocessor or any conventional processor.

[0176] Since the instructions stored in the storage medium can execute the steps in any of the methods provided in the embodiments of the present invention, the beneficial effects of any of the methods provided in the embodiments of the present invention can be achieved, as detailed in the preceding embodiments, and will not be repeated here.

[0177] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for visualizing and collaborating across the entire logistics chain, characterized in that, The method includes: Semantic parsing is performed on the logistics collaboration requirements input by the user to extract operation intent tags and key logistics entities. The operation intent tags are mapped to corresponding intent representation vectors. Match the logistics human-computer interaction interface to be rendered according to the operation intention tag, and construct a corresponding candidate information set for each field to be filled in the logistics human-computer interaction interface according to the logistics key entity and the pre-acquired logistics historical operation data. Based on the pre-stored logistics data, each candidate information in the candidate information set is evaluated and calculated to determine the first target candidate information that has the least impact on the change of other fields to be filled. The first target candidate information is used to initially fill the fields to be filled in the logistics human-computer interaction interface and to display the initial interface assembly. In response to a user's modification operation on the target field to be filled through the logistics human-computer interaction interface, the derived constraint boundary is generated by extracting the new content to be filled corresponding to the modification operation based on the logistics data dependency topology, and the affected field that conflicts with the derived constraint boundary is located. Based on the attribute dimension values ​​in the initial filling content of the affected field that do not conflict with the derived constraint boundary, retain attribute features are constructed. Candidate information that satisfies the derived constraint boundary is selected from the candidate information set corresponding to the affected field. The feature distance of the selected candidate information is calculated based on the preserved attribute features and the intent representation vector. The candidate information with the smallest feature distance is selected as the second target candidate information with the highest comprehensive matching degree, so as to drive the rendering node corresponding to the affected field to perform local data update and replacement display.

2. The method according to claim 1, characterized in that, The step of constructing a corresponding candidate information set for each field to be filled in the logistics human-computer interaction interface based on the key logistics entities and the pre-acquired historical logistics operation data specifically includes: Based on the key logistics entities, an initial candidate set is obtained by performing an association search on the historical logistics operation data. Acquire real-time status warning data of the current logistics network environment, and remove candidate information with abnormal attribute features that match the real-time status warning data from the initial candidate set to obtain a usable candidate set; Each candidate information in the available candidate set is converted into a business feature vector, and the feature distance between different candidate information in the available candidate set is calculated. Based on the feature distance, the candidate information in the available candidate set is sampled to generate the candidate information set that meets the preset capacity upper limit threshold. The feature distance between any two candidate information in the candidate information set is greater than the preset differentiation threshold.

3. The method according to claim 2, characterized in that, The step of evaluating and calculating each candidate information in the candidate information set based on the pre-stored logistics data dependency topology to determine the first target candidate information that has the least impact on the change magnitude of other fields to be filled specifically includes: Each candidate information in the candidate information set is sequentially used as a trigger node and input into the logistics data dependency topology to simulate constraint transmission, thereby extracting the cascading impact path that affects other fields to be filled. For each candidate information in the candidate information set, based on the cascading influence path, the number of options in the candidate information set that fail due to constraint conflicts for other fields to be filled affected by the candidate information is counted, and the number of options is summarized to calculate the space compression cost corresponding to the candidate information. The candidate information with the lowest spatial compression cost is determined as the first target candidate information.

4. The method according to claim 1, characterized in that, The step of extracting the derived constraint boundary corresponding to the new fill content of the modification operation based on the logistics data dependency topology specifically includes: Obtain the new fill content input to the target field to be filled corresponding to the modification operation; The newly filled content is compared with the original preliminary filled content of the target field to be filled to extract the logistics attribute bias that has changed. Using the target field to be filled as the baseline node, the downstream nodes that have a causal relationship with the baseline node are traversed in the logistics data dependency topology, and the business constraint rules attached to the downstream nodes are obtained. The business constraint rules include the business attribute dimensions corresponding to the downstream nodes. The logistics attribute bias is substituted into the business constraint rules to perform influence range extrapolation, generating the derived constraint boundary composed of multi-dimensional attribute constraints. The derived constraint boundary is used to limit the range of legal candidate data allowed for the downstream node after being affected by the modification operation. Each dimension in the multi-dimensional attribute constraints corresponds to an attribute dimension of the affected field.

5. The method according to claim 4, characterized in that, The steps of filtering candidate information that satisfies the derived constraint boundary from the candidate information set corresponding to the affected field, calculating the feature distance of the filtered candidate information based on the preserved attribute features and the intent representation vector, and selecting the candidate information with the smallest feature distance as the second target candidate information with the highest comprehensive matching degree to drive the rendering node corresponding to the affected field to perform local data update and replacement display, specifically include: Candidate information that does not conflict with the derived constraint boundary is extracted from the candidate information set corresponding to the affected field to obtain a verification candidate set; Based on the node-level characteristics of the logistics data dependency topology, the values ​​of each non-conflicting attribute dimension contained in the retained attribute characteristics are transformed into multi-dimensional vectorization to generate a baseline state vector. Each candidate information in the verification candidate set is simulated and substituted into the node position corresponding to the affected field in the logistics data dependency topology, and the candidate connected subgraph activated by the transmission of each candidate information along the logistics data dependency topology is extracted. The node attribute features of each candidate connected subgraph are encoded according to the baseline state vector to obtain the candidate state vector; Calculate the Euclidean distance between the baseline state vector and each of the candidate state vectors, and select the candidate information corresponding to the smallest Euclidean distance as the second target candidate information, so as to drive the rendering node corresponding to the affected field to perform local data update and replacement display.

6. The method according to claim 5, characterized in that, The step of performing multi-dimensional vectorization transformation on the values ​​of each non-conflicting attribute dimension contained in the retained attribute features based on the node-level characteristics of the logistics data dependency topology to generate a baseline state vector specifically includes: Extract the deviation direction between the new filling content and the initial filling content corresponding to the modification operation in the feature semantic space, and generate an intention drift vector by combining the logistics attribute bias. The intent representation vector and the business attribute dimension of the business constraint rule are jointly mapped to the same source representation space to generate an initial intent reference vector, and the feature distance between the intent drift vector and the initial intent reference vector is calculated to obtain the semantic discrete bias. When the semantic discrete bias is detected to be less than the preset redirection threshold, the baseline state vector is directly generated based on the retained attribute features; When the semantic discrete bias is detected to exceed the preset redirection threshold, the values ​​of each non-conflicting attribute dimension contained in the retained attribute feature are used to construct an initial attribute matrix. Obtain the adjacent association edge weights of the affected field in the logistics data dependency topology, and use the adjacent association edge weights to weight the values ​​of each dimension in the initial attribute matrix to generate a local feature weighted matrix; Based on the node-level characteristics of the logistics data dependency topology, the matrix parameters contained in the local feature weighting matrix are encoded using multidimensional vectorization to obtain the baseline state vector.

7. The method according to claim 1, characterized in that, After the steps of filtering candidate information that satisfies the derived constraint boundary from the candidate information set corresponding to the affected field, calculating the feature distance of the filtered candidate information based on the preserved attribute features and the intent representation vector, and selecting the candidate information with the smallest feature distance as the second target candidate information with the highest comprehensive matching degree to drive the rendering node corresponding to the affected field to perform local data update and replacement display, the method further includes: Extract the feature deviation between the initial filling content corresponding to the affected field and the second target candidate information, and perform a weighted calculation on the feature deviation with a preset business weight matrix to obtain the verification priority parameter. The business weight matrix is ​​composed of the business importance coefficients of each node in the logistics data dependency topology. When the verification priority parameter is greater than the preset threshold, the directed transmission path from the target field to be filled to the affected field in the logistics data dependency topology is mapped to the screen coordinate system of the logistics interaction interface to generate a page linkage guidance trajectory. The floating interface component is rendered based on the page linkage guide trajectory, and the initial filling content and the second target candidate information are output in parallel in the floating interface component to generate a change comparison view; In response to the update confirmation interaction command for the change comparison view within the scope of the page linkage guidance trajectory, the affected fields are updated and confirmed.

8. A logistics end-to-end visual collaborative system, characterized in that, The system includes: one or more processors and memory; The memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is run on the logistics end-to-end visual collaboration system, the system performs the method as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is run on the logistics end-to-end visualization and collaboration system, the system performs the method as described in any one of claims 1-7.

Images