Government affair item handling recommendation method and related device
By constructing a dynamic behavior graph to analyze the dependencies between government affairs, the problems of inaccurate user status and duplicate material submission in traditional government affairs recommendation systems have been solved, enabling personalized service recommendations and efficient handling of affairs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-05
- Publication Date
- 2026-04-07
AI Technical Summary
Traditional government service recommendation systems cannot effectively handle individual user differences and mandatory prerequisites between services, resulting in inaccurate recommendations and requiring users to submit materials repeatedly, thus reducing efficiency.
By acquiring user behavior data and government service rules, a dynamic behavior graph is constructed to analyze the dependencies between items, generate dynamic state tuples, determine the items that can be processed and their required materials, and make personalized recommendations based on the graph.
It enables personalized service recommendations based on user status, reducing the need for repetitive document submissions and improving efficiency and user experience.
Smart Images
Figure CN121810464A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, specifically to a recommended method and related apparatus for handling government affairs. Background Technology
[0002] Traditional recommendation systems are mostly based on static rules or general recommendation algorithms, such as collaborative filtering and content recommendation. However, these methods cannot meet the complex dependencies and dynamic processing procedures unique to government affairs. Currently, the mainstream approach to constructing government knowledge graphs mainly uses triples to represent the relationships between entities. This modeling approach has the following problems:
[0003] First, it cannot represent individual user differences; the same matter may be in different processing statuses for different users. Second, government service recommendations face unique challenges, as there are mandatory prerequisite dependencies between matters. For example, a birth certificate must be obtained before household registration can be processed, and household registration must be processed before an ID card can be obtained.
[0004] The relevant technologies cannot effectively handle these cascading dependencies to achieve personalized recommendations for matters that users can handle. Summary of the Invention
[0005] In view of this, this application provides a method and related device for recommending government affairs matters. In government service scenarios, it can provide users with more accurate recommendation results, guide users to handle the current matters that can be handled after handling the prerequisite matters, and at the same time reduce the probability of users submitting duplicate materials when handling the current matters that can be handled, thereby improving the efficiency of users.
[0006] To solve the above problems, the technical solution provided in this application is as follows:
[0007] On the one hand, this application provides a recommended method for handling government affairs, the method including:
[0008] Acquire the behavioral data of the first user and the government service rules, wherein the government service rules include the list of materials corresponding to different matters and the dependencies between the different matters;
[0009] Based on the behavioral data, a dynamic state tuple for the first user is generated. The dynamic state tuple is used to identify the state information corresponding to the first user. The dynamic state tuple includes at least the state of the first user handling the different matters.
[0010] Based on the government service rules and the dynamic state tuple, a dynamic behavior graph of the first user is constructed.
[0011] Based on the dynamic behavior graph, a list of materials to be provided corresponding to the matters that the first user can handle is determined.
[0012] Displays the list of materials to be provided for the matters that the first user can handle.
[0013] In one possible implementation, constructing the dynamic behavior graph of the first user based on the government service rules and the dynamic state tuple includes:
[0014] Based on the aforementioned government service rules, the dependencies between different items are analyzed, and a directed graph is constructed.
[0015] The dynamic state tuples are mapped to the directed graph to obtain the dynamic behavior graph corresponding to the first user.
[0016] In one possible implementation, determining the list of materials to be provided corresponding to the first user's eligible matters based on the dynamic behavior graph includes:
[0017] Based on the dynamic behavior graph, it is determined whether the prerequisite dependent items of the first item in different items are in the completed state. The first item is the item that the first user has not completed.
[0018] If the prerequisites of the first item are already processed, then the first item will be considered as the first user's processable item.
[0019] The list of materials required for processing the pursuant to ...
[0020] In one possible implementation, the number of pursuable items is multiple, and after generating the required materials corresponding to the pursuable items of the first user based on the dynamic behavior graph, the method further includes:
[0021] Based on the list of materials to be provided for each of the multiple items that can be processed, the priority of the multiple items that can be processed is determined;
[0022] The list of materials to be provided for the matters that can be handled by the first user includes:
[0023] Based on the priority, a recommendation result is displayed for the first user, which includes the multiple available items arranged in order and the corresponding list of materials to be submitted.
[0024] In one possible implementation, the method further includes:
[0025] Based on the dynamic behavior graph, the items that can be processed in parallel among the multiple processing items are determined. The items that can be processed in parallel are determined by the urgency of the items and / or the degree of material reuse.
[0026] The step of displaying the recommendation results for the first user based on the priority includes:
[0027] The recommended results and parallel processing prompts are displayed. The parallel processing prompts indicate that the first user is prompted to handle the items that can be handled in parallel.
[0028] In one possible implementation, the method further includes:
[0029] In response to the input operation of the first user query information, based on the dynamic behavior graph, a candidate intent is predicted. The candidate intent includes multiple candidate items that the first user intends to handle and the predicted probability of each candidate item among the multiple candidate items.
[0030] Confirm the semantic matching degree between the information corresponding to the input operation and each candidate item;
[0031] Based on the predicted probability of each candidate item in the candidate intent and the semantic matching degree, the target item of the first user intent is determined.
[0032] Display the target item.
[0033] On another front, this application provides a recommendation device for handling government affairs, the device comprising:
[0034] The acquisition unit is used to acquire the behavior data of the first user and the government service rules, wherein the government service rules include the material list for handling different matters and the dependency relationship between the different matters;
[0035] The generation unit is configured to generate a dynamic state tuple for the first user based on the behavioral data. The dynamic state tuple is used to identify the state information corresponding to the first user, and the dynamic state tuple includes at least the state of the first user handling the different matters.
[0036] The construction unit is used to construct the dynamic behavior graph of the first user based on the government service rules and the dynamic state tuple;
[0037] The determining unit is used to determine, based on the dynamic behavior graph, the list of materials to be provided corresponding to the matters that the first user can handle;
[0038] The display unit is used to display the list of materials to be provided for the matters that the first user can handle.
[0039] In one possible implementation, the building unit is used for:
[0040] Based on the aforementioned government service rules, the dependencies between different items are analyzed, and a directed graph is constructed.
[0041] The dynamic state tuples are mapped to the directed graph to obtain the dynamic behavior graph corresponding to the first user.
[0042] In one possible implementation, the determining unit is configured to:
[0043] Based on the dynamic behavior graph, it is determined whether the prerequisite dependent items of the first item in different items are in the completed state. The first item is the item that the first user has not completed.
[0044] If the prerequisites of the first item are already processed, then the first item will be considered as the first user's processable item.
[0045] The list of materials required for processing the pursuant to ...
[0046] In one possible implementation, the number of processable items is multiple, and the determining unit is further configured to:
[0047] Based on the list of materials to be provided for each of the multiple items that can be processed, the priority of the multiple items that can be processed is determined;
[0048] The display unit is used to display recommendation results for the first user based on the priority, and the recommendation results include the plurality of matters that can be handled in sequence and the corresponding list of materials to be submitted.
[0049] In one possible implementation, the determining unit is further configured to:
[0050] Based on the dynamic behavior graph, the items that can be processed in parallel among the multiple processing items are determined. The items that can be processed in parallel are determined by the urgency of the items and / or the degree of material reuse.
[0051] The display unit is used for:
[0052] The recommended results and parallel processing prompts are displayed. The parallel processing prompts indicate that the first user is prompted to handle the items that can be handled in parallel.
[0053] In one possible implementation, the apparatus further includes a prediction unit for:
[0054] In response to the input operation of the first user query information, based on the dynamic behavior graph, a candidate intent is predicted. The candidate intent includes multiple candidate items that the first user intends to handle and the predicted probability of each candidate item among the multiple candidate items.
[0055] The determining unit is further configured to:
[0056] Confirm the semantic matching degree between the information corresponding to the input operation and each candidate item;
[0057] Based on the predicted probability of each candidate item in the candidate intent and the semantic matching degree, the target item of the first user intent is determined.
[0058] The display unit is also used to display the target item.
[0059] In another aspect, this application provides a computer device, which includes a processor and a memory:
[0060] The memory is used to store computer programs;
[0061] The processor is configured to execute the method described in any of the above-described embodiments according to the computer program.
[0062] In another aspect, this application provides a computer-readable storage medium for storing a computer program that, when executed by a computer device, implements the method described in any of the above-mentioned embodiments.
[0063] In another aspect, this application provides a computer program product including a computer program, which, when run on a computer device, causes the computer device to perform any of the methods described above.
[0064] As can be seen from the above technical solution, this solution first acquires the behavioral data of the first user and the government service rules. Then, based on the behavioral data, it generates a dynamic state tuple corresponding to the first user, providing a rule-based description of the first user's dynamic behavior. This dynamic state tuple includes at least the status of the first user handling different matters. Next, based on the government service rules and the dynamic state tuple, it constructs a dynamic behavior graph of the first user, converting the static knowledge graph describing objective government service rules into a dynamic behavior graph describing the first user's dynamic behavior. Then, based on the dependencies between different matters in the dynamic behavior graph, it can infer the first user's current available matters based on the matters already handled and those not yet handled. Furthermore, based on the material lists corresponding to different matters, it determines and displays the list of materials to be provided for the first user's available matters. Therefore, in government service scenarios, more accurate and personalized recommendations for handling matters can be obtained based on the user's dynamic behavior, thereby guiding the user to handle available matters and reducing the probability of submitting duplicate materials when handling available matters. This addresses the unique issue of material reuse in government service scenarios and effectively improves user efficiency. Attached Figure Description
[0065] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0066] Figure 1 A flowchart illustrating a recommended method for handling government affairs, provided as an embodiment of this application;
[0067] Figure 2 This is a schematic diagram of a recommended device for handling government affairs, provided in an embodiment of this application. Detailed Implementation
[0068] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.
[0069] As described in the background section, current service recommendation methods in government service scenarios have the following problems:
[0070] (1) To-do items are out of the user's state: Traditional knowledge graphs use static triples, which cannot record dynamic information such as "what items the user has completed", resulting in recommended items being out of the user's real state and low recommendation accuracy.
[0071] (2) Lack of cascading reasoning mechanism: It is impossible to infer related subsequent matters based on completed matters, which often leads to the recommendation of irrelevant matters, increasing the user's operating costs and the system's error correction burden.
[0072] (3) Serious duplicate submission of materials: The inability to handle the unique "material reuse" relationship of government services (such as ID cards being used for different matters) leads to users submitting the same materials repeatedly, reducing efficiency and increasing the review pressure on government service windows.
[0073] To address the aforementioned issues in government service recommendation scenarios, this application provides a method for recommending government service items. By combining user behavior data and government service rules, a static knowledge graph describing objective government service rules is transformed into a dynamic behavior graph describing the dynamic behavior of the first user. Based on this dynamic behavior graph, more accurate and personalized recommendations for handling government service items can be obtained, thereby guiding users to process available items and reducing the probability of users submitting duplicate materials when processing available items. This addresses the issue of material reuse unique to government service scenarios and effectively improves user efficiency.
[0074] The solutions provided in this application relate to the field of computer technology, and are specifically illustrated through the following embodiments.
[0075] See Figure 1 The diagram shown is a flowchart illustrating a recommended method for handling government affairs according to an embodiment of this application. In this embodiment, it can be executed by a computer device.
[0076] S101: Obtain the behavior data of the first user and the rules of government services.
[0077] Among them, the rules for handling government services are used to identify the rules for handling different matters in government services. At least, they include the list of materials corresponding to handling different matters (such as the list of materials required for handling and the materials that can be obtained after successful handling), the dependencies between different matters, and may also include the materials that the client needs to provide when handling different matters, etc.
[0078] This application does not impose specific restrictions on the method of obtaining government service rules. For example, entities and relationships between entities are extracted from multi-source heterogeneous government data (such as unstructured or semi-structured data from government internal registration systems, policy documents, government websites, and user behavior logs). Then, using large-scale modeling or natural language processing techniques, the entity(s) of the matter, the entity(s) of the materials, and the logical relationships between them are extracted from the government data to determine the dependencies of different matters and the materials required to handle different matters. The dependencies can be determined as follows: if materials obtained by successfully handling matter B are required before handling matter A, then matter B is considered a mandatory prerequisite for matter A. If materials obtained by successfully handling matters D and E are required before handling matter C, and materials obtained from handling other matters are not required when handling matters D and E, then matters D and E are both mandatory prerequisites for matter A, and matters D and E are considered matters that can be processed in parallel. Then, the data is structured, and rule templates are constructed using predefined entity types and relationship types to obtain a static knowledge base for describing objective rules.
[0079] Behavioral data refers to the data generated when a user handles government affairs. This application embodiment does not impose specific restrictions on the specific content of behavioral data. For example, the behavioral data may include at least the status of the user handling different matters, and may also include the materials submitted for handling the matters, the materials obtained, the time corresponding to the successful handling of the matters, the user's identity information (applicant, agent), etc.
[0080] By leveraging government service rules, a solid data foundation can be provided for inferring what services a user can currently handle. Furthermore, by combining this with user behavior data, more accurate personalized inferences can be made for the user.
[0081] S102: Based on behavioral data, generate a dynamic state tuple corresponding to the first user.
[0082] The dynamic state tuple is used to identify the state information corresponding to the first user. This application embodiment does not impose specific limitations on the specific content of the dynamic state tuple. The dynamic state tuple at least includes the state of the first user handling different matters, and may also include a list of materials corresponding to the first user after successfully handling different matters, as well as a timestamp corresponding to the successful handling of the matters, etc.
[0083] This application provides a representation for dynamic state tuples. Specifically, it defines a novel knowledge representation unit, namely, a government knowledge quadruple, whose structure is defined as: Q = (User_ID, Item_ID, Relation, State_Vector). Here, User_ID is the unique identifier of the user, Item_ID is the unique identifier of the item, Relation represents the user's identity information (e.g., self, authorized representative, etc.), and different identity information requires different materials to handle the same item. State_Vector is a composite state vector used to describe in detail the state information of the user handling the item.
[0084] The status vector includes at least: completion_status, which is a boolean flag indicating whether the user has completed the task; and material_reuse_list, which is a data set used to record the list of materials generated after the task is completed and can be reused for other tasks. It may also include timestamp, which is used to record the specific time when the task was completed, providing a basis for time series analysis and process traceability.
[0085] Therefore, by constructing a quadruple as a dynamic state tuple to describe the user's dynamic behavior, the problem that traditional triples cannot accurately express the status of the user's actions can be solved.
[0086] This application embodiment can also use an attribute graph model to represent user behavior data. In the attribute graph, users and items can be used as nodes, and the user's action of handling an item can be represented as an edge with attributes, which can include information such as timestamps and reusable materials. This approach is more natively implemented in graph databases, but the quadruple model in this application embodiment logically treats the dynamic fact of "user-item relationship" as an independent knowledge unit, making the concept clearer and facilitating batch management of states and logical reasoning.
[0087] S103: Based on the government service rules and the dynamic state tuple, construct the dynamic behavior graph corresponding to the first user.
[0088] By combining the dynamic state tuple of the first user with the government service rules, a dynamic knowledge graph corresponding to the first user can be constructed. This application embodiment does not impose specific restrictions on the construction method of the dynamic knowledge graph.
[0089] For example, the dynamic state tuple is represented by a hierarchical structure by establishing a connection between the unique identifier of the item in the four-tuple corresponding to the first user and different items in the government service rules.
[0090] Therefore, by embedding dynamic state tuples corresponding to user behavior into government service rules, the static knowledge base used to describe objective government service rules can be transformed into a dynamic behavior graph. This dynamic behavior graph can reflect the progress of each user's work in real time. Thus, each user's dynamic behavior graph is a real-time snapshot of their personal work trajectory, which makes it possible to achieve truly personalized recommendations.
[0091] In one possible implementation, S103 includes:
[0092] A1: Based on the rules of government services, the dependencies between different items are analyzed and a directed graph is constructed.
[0093] A2: Map the dynamic state tuples to the directed graph to obtain the dynamic behavior graph corresponding to the first user.
[0094] This application's embodiments abstract and model the mandatory prerequisite dependencies between different items as a directed graph (DAG). In this directed graph, each node represents an item. If item A is a prerequisite item of item B, then there exists a directed edge from node A to node B.
[0095] Then, based on the relationships between the items, the dynamic state tuples are mapped to a directed graph to obtain a dynamic behavior graph. This ensures that the attributes of each node include at least the status of the first user handling the item, the corresponding list of materials (e.g., the list of materials to be submitted for handling the item, and the materials that can be obtained after successful handling), and may also include the timestamp corresponding to the successful handling of the item, etc.
[0096] Therefore, by forming an objective chain of dependencies through a directed graph, and then representing the behavior of different users with differentiated dynamic behavior graphs, the ability to cascade reasoning about currently actionable matters can be effectively improved, thereby enhancing reasoning efficiency.
[0097] S104: Based on the dynamic behavior graph, determine the list of materials to be provided for the first user's eligible matters.
[0098] The term "processable items" refers to items that a user can process. This application embodiment does not impose specific restrictions on the method of determining processable items. For example, if materials obtained after successfully processing item B are required before processing item A, then if item B is included in the first user's processed items but not item A, then item A is determined to be a processable item. Another example is when the dependency relationship is item A->item B->item C, item C is considered a processable item only if the processing status of item C corresponding to the first user is not processed, and both item A and item B have been successfully processed.
[0099] The list of materials to be provided refers to the list of materials that the first user still needs to provide when handling the matters that can be handled. Specifically, the list of materials corresponding to the matters that the user has already handled is compared with the list of materials corresponding to the matters that can be handled, and then the list of materials that the user still needs to provide for the matters that can be handled is determined.
[0100] Furthermore, through dynamic behavior graphs, it is possible to infer the matters that the first user can handle and the corresponding list of materials to be provided. By comparing the materials, it is possible to determine the list of materials that the first user still needs to provide, thus realizing a personalized reasoning method for handling government affairs.
[0101] In one possible implementation, S104 includes:
[0102] B1: Based on dynamic behavior graphs, determine whether the prerequisite dependent items of the first item in different items are in the "processed" state.
[0103] B2: If the prerequisites of the first item are already processed, then the first item will be considered as the first user's processable item.
[0104] B3: Compare the list of materials required for processing the eligible items with the list of materials corresponding to the items already processed by the first user to determine the list of materials to be provided for the eligible items.
[0105] Among them, the preceding dependent items refer to the preceding items that are directly related to the current item. For example, if item B can only be processed after item A is processed, then item A is a preceding dependent item of item B. Furthermore, there may be multiple preceding dependent items of item B. For example, if item A and item C are items that can be processed in parallel, and item B can only be processed after item A and item C are processed, then item A and item C are both preceding dependent items of item B.
[0106] The first item is the item that the first user is handling. If all the prerequisite items for the first item have been handled, then the first item is the item that the first user can handle. Then, based on the list of materials corresponding to handling different items, the materials that still need to be provided for handling the first item are determined.
[0107] This application embodiment, based on a dynamic behavior graph, designs a cascaded reasoning algorithm. The execution process of this algorithm is as follows: First, when a recommendation needs to be made for the first user, the "list of completed items" for the first user is obtained by querying the dynamic state tuple corresponding to the first user. Subsequently, all items in the government service rules are checked. For each item that has not yet been completed by the first user, its predecessor dependencies are checked. For example, in the DAG constructed based on the dependency relationship, all direct predecessor nodes corresponding to the first user's uncompleted items are determined. Only when all the start and end dependencies of an item exist in the first user's "list of completed items" is the item determined as an "applicable item". After generating the list of applicable items, material reuse optimization is further performed. For each "applicable item" in the list, its required material list is parsed and compared with the material list corresponding to the completed items to identify the material list that still needs to be provided by the first user, thus obtaining the material list to be provided corresponding to the applicable item.
[0108] Therefore, the first item is only considered as the first user's actionable item when all its prerequisites are already processed. This effectively reduces scenarios such as "repeated trips due to missing materials" and "failure to pass pre-qualification review," thereby further improving user efficiency.
[0109] S105: Displays the list of materials to be provided for the first user's eligible tasks.
[0110] This application embodiment does not impose specific restrictions on the display location of the list of materials to be provided corresponding to the first user's eligible matters. For example, it can be displayed on the homepage when the user logs into the government service webpage or APP, or below the query entry for the eligible matters, etc., so as to guide the user to handle the eligible matters.
[0111] Therefore, the system first acquires the behavioral data of the first user and the government service rules. Then, based on the behavioral data, it generates a dynamic state tuple corresponding to the first user, providing a rule-based description of the first user's dynamic behavior. This dynamic state tuple includes at least the status of the first user handling different matters. Next, based on the government service rules and the dynamic state tuple, it constructs a dynamic behavior graph of the first user, converting the static knowledge graph describing objective government service rules into a dynamic behavior graph describing the first user's dynamic behavior. Then, based on the dependencies between different matters in the dynamic behavior graph, it infers the first user's current available matters based on their completed and uncompleted matters. Furthermore, based on the material lists corresponding to different matters, it determines and displays the list of materials to be provided for the first user's available matters. Thus, in government service scenarios, more accurate and personalized recommendations for handling matters can be obtained based on the user's dynamic behavior, guiding the user to handle available matters and reducing the probability of submitting duplicate materials. This addresses the issue of material reuse unique to government service scenarios, effectively improving user efficiency.
[0112] This application also provides a method for predicting actionable items, such as using a graph neural network (GNN)-based predictive model for reasoning. By training the GNN model on a large amount of historical user behavior data, it learns the sequence patterns of item handling, thereby predicting the most likely next item a user will handle. The advantage of GNNs is their ability to discover hidden, non-explicit relationships in data, but their disadvantages include the need for massive amounts of labeled data and the fact that the model's reasoning process is a "black box," resulting in poor interpretability. For the government sector, which requires strong compliance and clear, transparent process logic, the deterministic reasoning scheme based on DAGs in this invention is more reliable and robust.
[0113] In one possible implementation, there are multiple pursuant to-do items, and the method further includes:
[0114] C1: Based on the list of materials to be provided for each of the multiple processable items, determine the priority of the multiple processable items.
[0115] S105 includes:
[0116] C2: Based on priority, it displays the recommended results for the first user. The recommended results include multiple items that can be processed in order and the corresponding list of materials to be submitted.
[0117] For example, the priority of multiple processable items can be determined based on the number of materials to be provided for each item or the difficulty of processing the materials to be provided. For example, if there is only one material in the materials to be provided for item A among multiple items, item A can be given a higher priority. Or, if the materials included in the materials to be provided for item A require a longer time to obtain accurate information (such as verification), a lower priority can be set for them.
[0118] Then, based on priority, the system displays recommendations for the first user. These recommendations include a list of available tasks displayed in order of priority, along with a corresponding list of required documents. For example, higher-priority tasks are listed first to allow users to prioritize easier tasks.
[0119] Therefore, by using the list of materials to be provided, the priority of the matters that can be handled can be determined, and then the matters with higher priority can be displayed on the recommendation page. This can achieve a precise matching from "people looking for services" to "services looking for people", providing users with more accurate and personalized recommendation results.
[0120] In one possible implementation, the method further includes:
[0121] D1: Based on the dynamic behavior graph, identify the items that can be processed in parallel among multiple processing items.
[0122] C2 includes:
[0123] D2: Displays recommended results and parallel processing prompts.
[0124] The parallel processing prompt indicates that the first user can handle tasks that can be handled in parallel. This application embodiment does not impose specific limitations on this; for example, it may also include prompts regarding the urgency of the task or the degree of material reuse.
[0125] Among them, the items that can be processed in parallel are determined based on the urgency of the items and / or the degree of material reuse. For example, multiple items that meet the preset urgency conditions (such as the processing time of items A and B being longer than the preset time) are considered as items that can be processed in parallel, and multiple items that meet the preset standard of material reuse (such as the overlap of the list of materials to be provided reaching 80%) are considered as items that can be processed in parallel. Alternatively, multiple items that can be processed in parallel can be determined by weighted merging based on urgency and material reuse.
[0126] Then, multiple items that can be processed and their corresponding lists of required materials are displayed in order of priority, and a parallel processing prompt is displayed at the same time. For example, the phrase "Processing time is long, parallel processing is possible" is marked after items A and B that can be processed in parallel, or the user is prompted by displaying "The overlap of required materials for items A and B is more than 80%, it is recommended to process them in parallel".
[0127] Therefore, by providing parallel processing prompts, users can be advised to process tasks in parallel based on their urgency or the reusability of materials, thereby further helping them improve their efficiency.
[0128] This application also provides a method for generating recommendation results, which is illustrated by the following steps.
[0129] S1: Get the set of completed items for the first user;
[0130] S2: Construct a directed acyclic graph based on the dependencies between items, where nodes represent items and edges represent predecessor dependencies;
[0131] S3: Perform topological sorting on the DAG, and sequentially determine whether the prerequisites of each unprocessed item are in a processed state. If so, add them to the "List of Items that Can Be Processed".
[0132] S4: Conduct material reuse analysis on the matters that can be processed, identify reusable materials, and determine the corresponding list of materials to be submitted;
[0133] S5: Calculate priority scores based on factors such as the degree of material reuse and the urgency of the matter, and output recommendations in order of scores.
[0134] This application provides a method for intent recognition. In one possible implementation, the method further includes:
[0135] E1: In response to the input operation of the first user query information, predict candidate intents based on dynamic behavior graph.
[0136] E2: Confirm the semantic matching degree between the information corresponding to the input operation and each candidate item.
[0137] E3: Based on the predicted probability and semantic matching degree of each candidate item in the candidate intent, determine the target item to be handled by the first user intent.
[0138] E4: Displays the target item.
[0139] Among them, the candidate intent includes multiple candidate items that the first user intends to handle, as well as the predicted probability of each candidate item.
[0140] The target matter refers to the matter that the first user intends to handle, as determined by the information input by the first user. It may include one or more, and this application embodiment does not impose specific limitations on it.
[0141] For example, when the first user enters a fuzzy query term (such as "obtain a certificate" or "start a company"), the system no longer simply performs keyword matching, but instead performs an intent completion inference process based on a dynamic behavior graph.
[0142] The process first obtains one or more recently completed tasks from the first user based on the timestamps of completed tasks, forming their current "task chain". Then, using a Directed Acyclic Graph (DAG), starting from the end node of this task chain, it predicts all direct subsequent tasks or subsequent task paths to obtain a high-probability, narrow-range candidate intent. This can be achieved using Markov chains, or pre-built LSTM, graph neural networks, etc., to determine multiple candidate tasks that the first user intends to handle and their corresponding predicted probabilities.
[0143] Then, a semantic similarity calculation model is used to analyze the degree of matching between the fuzzy query terms input by the first user and the text descriptions of each item in the candidate intent set. Finally, by weighted fusion of the predicted probabilities of each candidate item in the candidate intent and the semantic matching degree of the query text, a final score is obtained. This score is used to determine the first user's target item. If the final score is greater than a preset threshold, the candidate item is selected as the target item. Then, the target item or its processing entry is displayed in descending order of the final score.
[0144] The following steps illustrate the method of intent recognition.
[0145] S11: Obtain several recently completed items from the first user and arrange them into a "behavior chain" in reverse chronological order;
[0146] S12: Based on this behavioral chain, use sequence prediction models (such as Markov chains, LSTM, graph neural networks) to predict several candidate items that the user is most likely to handle next and their corresponding prediction probabilities, and obtain candidate intentions;
[0147] S13: Perform semantic similarity matching between the natural language query information input by the first user and several candidate items output by S2, and calculate the semantic matching score of each candidate item.
[0148] S14: The predicted probability of the behavior chain and the semantic matching score are weighted and fused according to the preset weights to generate the final recommended ranking list and return it.
[0149] Therefore, by combining the historical behavioral context of the first user with the semantics of the current query, the accuracy of intent recognition for fuzzy queries can be greatly improved, thereby effectively improving the accuracy of recommendation results.
[0150] This application also provides a method for intent recognition, specifically, a scheme based on a large-scale pre-trained language model (LLM). By taking "user's historical action chain + fuzzy query" as input, a large language model is fine-tuned to directly generate or classify the user's precise intent. LLM has powerful semantic understanding capabilities, but its deployment and inference costs are high, and its performance may not be as stable and accurate as the scheme proposed in this invention, which combines dynamic behavior graphs (DAG) and semantic matching, when dealing with specific domains and low-frequency long-tail items.
[0151] The recommended method for handling government affairs provided in this application embodiment can be deployed in a layered system architecture. For example, the architecture provided in this application embodiment may include a data layer, an engine layer, a service layer, and an application layer.
[0152] Data Layer: Responsible for the storage and management of all data. This includes a database for storing raw government data, a static knowledge base for storing global policy-dependent DAGs, and a dynamic state library for storing users' dynamic state quadruples.
[0153] Engine Layer: This is the core logic processing center of the system. It includes the "quadruple construction and update module" responsible for generating and updating quadruples from raw data in real time, the "cascaded inference engine" that implements cascaded recommendation logic, and the "intent completion engine" that handles fuzzy user queries.
[0154] Service layer: Through standardized application programming interfaces (APIs), the core functions of the engine layer (such as recommendation generation and status query) are encapsulated into services for upper-layer applications to call.
[0155] Application layer: This is the terminal directly facing the user, such as government service apps, government portals, and intelligent customer service robots. They provide users with the final intelligent and personalized government services by calling the interfaces of the service layer.
[0156] Based on the above system architecture, the implementation steps of the technical solution of this application are illustrated by example.
[0157] Step 1: Constructing the quadruple.
[0158] Extract user processing records from the government system, construct a global dependency DAG, generate an initial set of four tuples and store them in a dynamic state database.
[0159] Step 2: Real-time updates.
[0160] Each time a user completes a task, an event listener is triggered; the corresponding four-tuple's State_Vector is updated (e.g., completion_status=true, timestamp=now); the cascading inference engine is automatically triggered to update the user's "list of tasks available for completion".
[0161] Step 3: Recommendation generation.
[0162] Generate a valid recommendation sequence based on the DAG topology; apply material reuse rules to mark "materials exempt from submission"; output the results in priority order and highlight the reasons for recommendation on the front end.
[0163] Therefore, the technical solution applied for has the following beneficial effects:
[0164] (1) It realizes personalized service from "one-size-fits-all" to "one-person-one-policy". By introducing dynamic state quadruple, the system can accurately track the real-time progress of each user's business. The recommendation results are generated entirely based on the user's personal status, which solves the problem of homogeneous and untargeted recommendation content in traditional systems and provides users with highly personalized business navigation.
[0165] (2) Improve the compliance and completeness of the recommendations. Based on the cascading reasoning algorithm of the DAG of the matters, the algorithm follows the legal procedures and logical order of government affairs, so that the matters recommended to users are legal and compliant in procedure. At the same time, the algorithm can automatically plan a complete subsequent service path for users, avoiding human oversights or errors caused by information asymmetry.
[0166] (3) Improve the efficiency of government services and user experience. By integrating features such as material reuse mechanisms into the reasoning process, the system can proactively and intelligently prompt users to reuse submitted materials, reduce the burden of users repeatedly preparing and submitting materials, simplify procedures, and improve the service experience for the public and businesses.
[0167] (4) Enhance the intelligence and naturalness of human-computer interaction. Based on the intent completion mechanism of the DAG based on the item dependency, the system can understand the user's fuzzy instructions in combination with the context, which improves the accuracy of search and question answering. This transforms the system from a passive information query tool into an intelligent service assistant that can proactively predict and respond accurately.
[0168] Based on the above embodiments, this application provides a recommendation device for handling government affairs, with reference to... Figure 2 The diagram shown is a schematic of a recommendation device for handling government affairs provided in an embodiment of this application. The device 200 includes:
[0169] The acquisition unit 201 is used to acquire the behavior data of the first user and the government service rules, wherein the government service rules include the material list for handling different matters and the dependency relationship between the different matters.
[0170] The generation unit 202 is used to generate a dynamic state tuple for the first user based on the behavior data. The dynamic state tuple is used to identify the state information corresponding to the first user. The dynamic state tuple includes at least the state of the first user handling the different matters.
[0171] Construction unit 203 is used to construct a dynamic behavior graph of the first user based on the government service rules and the dynamic state tuple;
[0172] The determining unit 204 is used to determine the list of materials to be provided corresponding to the matters that can be handled by the first user based on the dynamic behavior graph;
[0173] Display unit 205 is used to display the list of materials to be provided for the matters that the first user can handle.
[0174] In one possible implementation, the building unit is used for:
[0175] Based on the aforementioned government service rules, the dependencies between different items are analyzed, and a directed graph is constructed.
[0176] The dynamic state tuples are mapped to the directed graph to obtain the dynamic behavior graph corresponding to the first user.
[0177] In one possible implementation, the determining unit is configured to:
[0178] Based on the dynamic behavior graph, it is determined whether the prerequisite dependent items of the first item in different items are in the completed state. The first item is the item that the first user has not completed.
[0179] If the prerequisites of the first item are already processed, then the first item will be considered as the first user's processable item.
[0180] The list of materials required for processing the pursuant to ...
[0181] In one possible implementation, the number of processable items is multiple, and the determining unit is further configured to:
[0182] Based on the list of materials to be provided for each of the multiple items that can be processed, the priority of the multiple items that can be processed is determined;
[0183] The display unit is used to display recommendation results for the first user based on the priority, and the recommendation results include the plurality of matters that can be handled in sequence and the corresponding list of materials to be submitted.
[0184] In one possible implementation, the determining unit is further configured to:
[0185] Based on the dynamic behavior graph, the items that can be processed in parallel among the multiple processing items are determined. The items that can be processed in parallel are determined by the urgency of the items and / or the degree of material reuse.
[0186] The display unit is used for:
[0187] The recommended results and parallel processing prompts are displayed. The parallel processing prompts indicate that the first user is prompted to handle the items that can be handled in parallel.
[0188] In one possible implementation, the apparatus further includes a prediction unit for:
[0189] In response to the input operation of the first user query information, based on the dynamic behavior graph, a candidate intent is predicted. The candidate intent includes multiple candidate items that the first user intends to handle and the predicted probability of each candidate item among the multiple candidate items.
[0190] The determining unit is further configured to:
[0191] Confirm the semantic matching degree between the information corresponding to the input operation and each candidate item;
[0192] Based on the predicted probability of each candidate item in the candidate intent and the semantic matching degree, the target item of the first user intent is determined.
[0193] The display unit is also used to display the target item.
[0194] Based on the above embodiments, this application provides a computer device, which includes a processor and a memory:
[0195] The memory is used to store computer programs;
[0196] The processor is used to execute the recommended method for handling the aforementioned government affairs according to the computer program.
[0197] Based on the above embodiments, this application provides a computer-readable storage medium for storing a computer program that, when executed by a computer device, implements the recommended method for handling the aforementioned government affairs.
[0198] Based on the above embodiments, this application provides a computer program product including a computer program, which, when run on a computer device, causes the computer device to perform the recommended method for handling government affairs.
[0199] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems or apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and relevant parts can be referred to the method section.
[0200] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A recommended method for handling government affairs, characterized in that, The method includes: Acquire the behavioral data of the first user and the government service rules, wherein the government service rules include the list of materials corresponding to different matters and the dependencies between the different matters; Based on the behavioral data, a dynamic state tuple for the first user is generated. The dynamic state tuple is used to identify the state information corresponding to the first user. The dynamic state tuple includes at least the state of the first user handling the different matters. Based on the government service rules and the dynamic state tuple, a dynamic behavior graph of the first user is constructed. Based on the dynamic behavior graph, a list of materials to be provided corresponding to the matters that the first user can handle is determined. Displays the list of materials to be provided for the matters that the first user can handle.
2. The method according to claim 1, characterized in that, The construction of the dynamic behavior graph of the first user based on the government service rules and the dynamic state tuple includes: Based on the aforementioned government service rules, the dependencies between different items are analyzed, and a directed graph is constructed. The dynamic state tuples are mapped to the directed graph to obtain the dynamic behavior graph corresponding to the first user.
3. The method according to any one of claims 1-2, characterized in that, The step of determining the list of materials to be provided for the first user's eligible matters based on the dynamic behavior graph includes: Based on the dynamic behavior graph, it is determined whether the prerequisite dependent items of the first item in different items are in the completed state. The first item is the item that the first user has not completed. If the prerequisites of the first item are already processed, then the first item will be considered as the first user's processable item. The list of materials required for processing the pursuant to ...
4. The method according to claim 1, characterized in that, The number of pursuable items is multiple. After generating the required materials corresponding to the pursuable items of the first user based on the dynamic behavior graph, the method further includes: Based on the list of materials to be provided for each of the multiple items that can be processed, the priority of the multiple items that can be processed is determined; The list of materials to be provided for the matters that can be handled by the first user includes: Based on the priority, a recommendation result is displayed for the first user, which includes the multiple available items arranged in order and the corresponding list of materials to be submitted.
5. The method according to claim 4, characterized in that, The method further includes: Based on the dynamic behavior graph, the items that can be processed in parallel among the multiple processing items are determined. The items that can be processed in parallel are determined by the urgency of the items and / or the degree of material reuse. The step of displaying the recommendation results for the first user based on the priority includes: The recommended results and parallel processing prompts are displayed. The parallel processing prompts indicate that the first user is prompted to handle the items that can be handled in parallel.
6. The method according to claim 1, characterized in that, The method further includes: In response to the input operation of the first user query information, based on the dynamic behavior graph, a candidate intent is predicted. The candidate intent includes multiple candidate items that the first user intends to handle and the predicted probability of each candidate item among the multiple candidate items. Confirm the semantic matching degree between the information corresponding to the input operation and each candidate item; Based on the predicted probability of each candidate item in the candidate intent and the semantic matching degree, the target item of the first user intent is determined. Display the target item.
7. A recommendation device for handling government affairs, characterized in that, The device includes: The acquisition unit is used to acquire the behavior data of the first user and the government service rules, wherein the government service rules include the material list for handling different matters and the dependency relationship between the different matters; The generation unit is configured to generate a dynamic state tuple for the first user based on the behavioral data. The dynamic state tuple is used to identify the state information corresponding to the first user, and the dynamic state tuple includes at least the state of the first user handling the different matters. The construction unit is used to construct the dynamic behavior graph of the first user based on the government service rules and the dynamic state tuple; The determining unit is used to determine, based on the dynamic behavior graph, the list of materials to be provided corresponding to the matters that the first user can handle; The display unit is used to display the list of materials to be provided for the matters that the first user can handle.
8. A computer device, characterized in that, The computer device includes a processor and memory: The memory is used to store computer programs; The processor is configured to perform the method according to any one of claims 1-6 according to the computer program.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program that, when executed by a computer device, performs the method described in any one of claims 1-6.
10. A computer program product comprising a computer program, characterized in that, When it is run on a computer device, it causes the computer device to perform the method described in any one of claims 1-6.