Government affair data processing method and device and electronic equipment

Abstract

The application discloses a government affair data processing method and device and electronic equipment. The method comprises the following steps: receiving government affair data input by a user; in the case that the government affair data comprises a question sequence corresponding to a government affair analysis scene and a government affair data table, determining a classification result corresponding to the question sequence, and extracting target feature data in the government affair data table, wherein the target feature data only comprises a file table header of the government affair data table and standardized data of a preset data volume; determining a query result corresponding to the target feature data, and determining a government affair analysis result corresponding to the government affair data according to the query result and the classification result of the question sequence. The application solves the technical problem that the data processing efficiency of the government affair scene is poor due to the fact that the intelligent government affair system in the related art has resource and privacy risks in data processing, and cannot effectively process table data exceeding the token length limit.

Description

Technical Field

[0001] This application relates to the field of government services, and more specifically, to a government data processing method, apparatus, and electronic device. Background Technology

[0002] In the process of digitalizing government services, traditional manual processing methods are slow and inefficient. To improve office efficiency, some intelligent government systems have introduced large language models to realize functions such as data analysis, document generation and distribution.

[0003] However, in intelligent analysis scenarios, especially when processing structured government data (such as Excel, CSV and other tabular files), intelligent government systems in related technologies generally adopt the strategy of uploading the entire data to a large model for parsing. Due to the token length limit of a single request from the large model, massive amounts of tabular data need to be cut into multiple fragments for batch transmission and processing, resulting in a surge in computing resource consumption, significant processing delays, and the risk of leakage of sensitive government data during multiple uploads.

[0004] The full data upload results in huge computational resource consumption and increased risk of privacy leakage. Furthermore, large models cannot efficiently complete analysis tasks when processing tabular data exceeding token limits, leading to low overall response efficiency and insufficient accuracy. This severely restricts the practical value and scalability of intelligent government systems.

[0005] There is currently no effective solution to the above problems. Summary of the Invention

[0006] This application provides a government data processing method, apparatus, and electronic device to at least solve the technical problem that the data processing efficiency in government scenarios is poor due to resource and privacy risks in data processing of intelligent government systems in related technologies, and the inability to effectively process table data exceeding the token length limit.

[0007] According to one aspect of the embodiments of this application, a government data processing method is provided, comprising: receiving government data input by a user; in the case that the government data includes a problem sequence and a government data table corresponding to a government analysis scenario, determining a classification result corresponding to the problem sequence, and extracting target feature data from the government data table, wherein the classification result is used to indicate the problem type of the problem sequence, and the target feature data only includes the file header and standardized data of a preset amount from the government data table; determining a query result corresponding to the target feature data, and determining a government analysis result corresponding to the government data based on the query result and the classification result of the problem sequence.

[0008] Optionally, determining the classification result corresponding to the question sequence includes: generating a first semantic vector corresponding to the question sequence through a semantic vector model; extracting structured features from the question sequence, wherein the structured features include at least the question length and keyword frequency of the question sequence; determining a joint semantic vector corresponding to the question sequence based on the first semantic vector and the structured features; processing the joint semantic vector through a classifier to output a probability value corresponding to the question sequence, wherein the probability value is used to indicate the probability that the question sequence belongs to different question types; if the probability value is greater than or equal to a preset probability threshold, determining the classification result corresponding to the question sequence as a direct retrieval type, and if the probability value is less than the preset probability threshold, determining the classification result corresponding to the question sequence as a summary report type.

[0009] Optionally, the government affairs analysis results corresponding to the government affairs data are determined based on the query results and the classification results of the question sequence, including: if the classification results corresponding to the question sequence are of the direct retrieval type, the query results are determined as the government affairs analysis results corresponding to the government affairs data; if the classification results corresponding to the question sequence are of the summary report type, the processing logic template corresponding to the question sequence is determined, and the question sequence and processing logic template are analyzed through a data processing model to obtain a summary report corresponding to the question sequence, and the summary report is determined as the government affairs analysis results corresponding to the government affairs data.

[0010] Optionally, the method further includes: when the government data contains meeting data corresponding to the scenario of drafting government meeting documents, matching the meeting data with the local meeting document knowledge base to obtain a matching result; when the matching result indicates that the meeting data does not match the meeting document knowledge base, directly outputting the meeting document result corresponding to the meeting data through a preset interaction model; when the matching result indicates that the meeting data matches the meeting document knowledge base, determining the reference document with the highest similarity to the meeting data from the meeting document knowledge base, and determining the meeting document result corresponding to the meeting data based on the reference document.

[0011] Optionally, the reference document with the highest similarity to the meeting data is determined from the meeting document knowledge base, including: generating a second semantic vector corresponding to the meeting data through a semantic vector model; determining a first similarity between the second semantic vector and a third semantic vector in the meeting document knowledge base, wherein the third semantic vector is the semantic vector corresponding to any document template in the meeting document knowledge base; determining a first similarity set, and determining the document template corresponding to the highest similarity in the first similarity set as the reference document corresponding to the meeting data, wherein the first similarity set is the similarity set with a first similarity greater than a first preset screening threshold.

[0012] Optionally, the method further includes: receiving a first revision of the meeting draft results detected through a visual interface; and updating the meeting draft knowledge base based on the first revision.

[0013] Optionally, the method further includes: when the government data contains document data corresponding to the government document distribution scenario, processing the document data through a preset document decision model to obtain a distribution decision corresponding to the document data, wherein the distribution decision includes at least the initial document distribution unit, the initial document distribution content, and the confidence level corresponding to the document data; when the confidence level is greater than or equal to a preset confidence threshold, determining the initial document distribution unit and the initial document distribution content as the document distribution result corresponding to the document data; when the confidence level is less than the preset confidence threshold, determining the document distribution template with the highest similarity to the document data from the local document distribution knowledge base, and determining the target document distribution unit and the target document distribution content in the document distribution template as the document distribution result corresponding to the document data.

[0014] Optionally, the document distribution template with the highest similarity to the document data is determined from the local document distribution knowledge base, including: generating a fourth semantic vector corresponding to the document data through a semantic vector model; determining a second similarity between the fourth semantic vector and a fifth semantic vector in the document distribution database, wherein the fifth semantic vector is the semantic vector corresponding to any document distribution template in the document distribution knowledge base; determining a second similarity set, and determining the document distribution template corresponding to the maximum similarity in the second similarity set as the document distribution template corresponding to the document data, wherein the second similarity set is a similarity set with a second similarity greater than a second preset screening threshold.

[0015] Optionally, the method further includes: receiving a second revision of the document distribution result detected through a visual interface; and updating the document distribution knowledge base based on the second revision.

[0016] According to another aspect of the embodiments of this application, a government data processing apparatus is also provided, comprising: a receiving module for receiving government data input by a user; a first determining module for determining a classification result corresponding to the problem sequence and extracting target feature data from the government data table when the government data contains a problem sequence and a government data table corresponding to a government analysis scenario, wherein the classification result is used to indicate the problem type of the problem sequence, and the target feature data only contains the file header and standardized data of a preset amount from the government data table; and a second determining module for determining a query result corresponding to the target feature data and determining a government analysis result corresponding to the government data based on the query result and the classification result of the problem sequence.

[0017] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory and a processor, wherein the memory is used to store program instructions; and the processor is connected to the memory and used to execute the above-described government data processing method.

[0018] According to another aspect of the embodiments of this application, a non-volatile storage medium is also provided, the non-volatile storage medium including a stored computer program, wherein the device where the non-volatile storage medium is located executes the above-mentioned government data processing method by running the computer program.

[0019] According to another aspect of the embodiments of this application, a computer program product is also provided, including computer instructions that, when executed by a processor, implement the above-described government data processing method.

[0020] In this embodiment, by receiving government data input by a user; when the government data contains a question sequence and a government data table corresponding to the government analysis scenario, determining the classification result corresponding to the question sequence, and extracting target feature data from the government data table, wherein the classification result is used to indicate the question type of the question sequence, and the target feature data only contains the file header and standardized data of a preset amount from the government data table; determining the query result corresponding to the target feature data, and determining the government analysis result corresponding to the government data based on the query result and the classification result of the question sequence, the purpose of efficient government data processing through lightweight feature extraction and intelligent classification is achieved. This achieves the technical effect of reducing the resource consumption of large models and the risk of privacy leakage, and solves the technical problem that the intelligent government system in related technologies has resource and privacy risks in data processing and cannot effectively process table data exceeding the token length limit, resulting in poor data processing efficiency in government scenarios. Attached Figure Description

[0021] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0022] Figure 1 This is a hardware structure diagram of a computer terminal for implementing a government data processing method according to an embodiment of this application;

[0023] Figure 2 This is a flowchart of a government data processing method according to an embodiment of this application;

[0024] Figure 3 This is a data processing flowchart of an intelligent government affairs system in an intelligent analysis scenario according to an embodiment of this application;

[0025] Figure 4 This is a data processing flowchart of an intelligent government affairs system in an intelligent meeting drafting scenario according to an embodiment of this application;

[0026] Figure 5 This is a data processing flowchart of an intelligent government affairs system in a document distribution scenario according to an embodiment of this application;

[0027] Figure 6 This is a structural diagram of a government data processing device according to an embodiment of this application. Detailed Implementation

[0028] 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 skilled in the art without creative effort should fall within the scope of protection of the present application.

[0029] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] First, some nouns or terms that appear in the explanation of the embodiments of this application shall be interpreted as follows:

[0031] Token Length: This refers to the maximum number of units that a large language model can process in a single request for input and output text sequences. The text is divided into basic semantic units (called tokens), such as words, subwords, or punctuation marks. The model's processing capacity is limited by the total number of tokens that its context window can hold.

[0032] PandasAI: An intelligent data analysis library that adds generative AI capabilities to the popular Python data analysis library Pandas, allowing users to directly query, analyze, and visualize tabular data using natural language commands without writing complex code.

[0033] BGE-M3 (Bidirectional Encoder from General Embedding Model version 3): A general semantic embedding model that integrates dense retrieval, sparse retrieval, and multi-vector retrieval capabilities. It can generate high-quality, multi-dimensional semantic vector representations of text and is suitable for various NLP tasks such as retrieval, classification, and re-ranking.

[0034] Vector cosine similarity: a mathematical method to measure the degree of directional similarity between two vectors in a multidimensional space. It evaluates the closeness of their semantics or features by calculating the cosine value of the angle between the two vectors. The value ranges from [-1, 1]. The closer the value is to 1, the more consistent the directions are and the more similar the semantics are.

[0035] Beyond intelligent analytics scenarios, intelligent e-government systems in related technologies also exhibit significant shortcomings in intelligent meeting organization and document drafting, as well as document distribution. For instance, in intelligent document processing and meeting organization scenarios, intelligent e-government systems typically rely directly on large models to generate document content, lacking deep and dynamic integration with local knowledge bases (such as historical documents and policies / regulations). The generated results often deviate from standardized e-government terminology and format requirements, resulting in a high rate of manual correction. Furthermore, the static and fixed knowledge base cannot evolve in real-time with policy updates, leading to low reusability. In document distribution scenarios, intelligent e-government systems mostly rely on rule matching or manual judgment, failing to effectively utilize historical distribution cases for semantic association. This results in frequent errors in distribution and misdelivery of content. Moreover, local knowledge base updates depend on manual maintenance, making it difficult to adapt to the dynamic and personalized needs of e-government operations.

[0036] To address the aforementioned problems, this application provides a method for processing government data, which can be run on... Figure 1 The computer terminal shown is described below.

[0037] The government data processing method embodiments provided in this application can be executed on mobile terminals, computer terminals or similar computing devices. Figure 1 A hardware structure block diagram of a computer terminal for implementing a government data processing method is shown. Figure 1As shown, the computer terminal 10 may include one or more processors (shown as 102a, 102b, ..., 102n in the figure) (the processor may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission module 106 for communication functions connected via wired and / or wireless networks. In addition, it may also include: a display, a keyboard, a cursor control device, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, and a BUS bus. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0038] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10. As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0039] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the government data processing method in this embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned government data processing method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0040] The transmission module 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission module 106 includes a network interface controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission module 106 may be a radio frequency (RF) module, used for wireless communication with the Internet.

[0041] The display can be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10.

[0042] It should be noted here that, in some optional embodiments, the above... Figure 1 The computer terminal shown may include hardware elements (including circuitry), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware and software elements. It should be noted that... Figure 1 This is only one instance of a specific particular instance, and is intended to illustrate the types of components that may exist in the aforementioned computer terminal.

[0043] In the above operating environment, this application provides an embodiment of a government data processing method. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Also, although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than that shown here.

[0044] Figure 2 This is a flowchart of a government data processing method according to an embodiment of this application, such as... Figure 2 As shown, the method includes the following steps:

[0045] Step S202: Receive government data input by the user.

[0046] Step S204: If the government data contains a problem sequence and a government data table corresponding to the government analysis scenario, determine the classification result corresponding to the problem sequence and extract the target feature data from the government data table. The classification result is used to indicate the problem type of the problem sequence, and the target feature data only contains the file header of the government data table and standardized data of a preset amount.

[0047] Step S206: Determine the query results corresponding to the target feature data, and determine the government affairs analysis results corresponding to the government affairs data based on the query results and the classification results of the question sequence.

[0048] Through steps S202 to S206, the goal of efficient government data processing through lightweight feature extraction and intelligent classification is achieved. This reduces the resource consumption of large models and the risk of privacy leaks, thereby solving the technical problem of poor data processing efficiency in government scenarios caused by resource and privacy risks in data processing and the inability to effectively handle table data exceeding token length limits in related intelligent government systems. The details are as follows.

[0049] Figure 3 This is a data processing flowchart of an intelligent government affairs system in an intelligent analysis scenario according to an embodiment of this application. Corresponding to the data processing logic in steps S202 to S206 above, it specifically includes the following process:

[0050] 1. Classification of user issues.

[0051] First, the system receives government data input by the user. This means that the system obtains the original information related to government services submitted by the user through a preset interface or interactive interface. This information includes, but is not limited to, the question sequence and government data table corresponding to the intelligent analysis scenario (taking the government analysis scenario as an example in this application), the meeting data corresponding to the intelligent meeting and document drafting scenario (taking the government meeting and document drafting scenario as an example in this application), and the document data corresponding to the document distribution scenario. This information exists in a structured or unstructured form and is used as the input basis for subsequent processing.

[0052] In the context of government affairs analysis, for a sequence of questions contained in government affairs data, the corresponding classification result is determined, including: generating a first semantic vector corresponding to the question sequence through a semantic vector model; extracting structured features from the question sequence, wherein the structured features include at least the question length and keyword frequency of the question sequence; determining a joint semantic vector corresponding to the question sequence based on the first semantic vector and the structured features; processing the joint semantic vector through a classifier to output a probability value corresponding to the question sequence, wherein the probability value is used to indicate the probability that the question sequence belongs to different question types; if the probability value is greater than or equal to a preset probability threshold, the classification result corresponding to the question sequence is determined as direct retrieval type, and if the probability value is less than the preset probability threshold, the classification result corresponding to the question sequence is determined as summary report type.

[0053] For example: Suppose the user input is a natural language sequence. First, a sequence of questions is generated using a pre-trained language model or the BGE-M3 model (i.e., the semantic vector model mentioned above). The corresponding first semantic vector and integrate the question length Construct a joint semantic vector using structured features such as keyword frequency. Subsequently, based on probability threshold decisions, a lightweight classifier outputs the corresponding probability values. And according to the preset probability threshold , the problem sequence Mapped to a binary category space, it can be represented as ,like The problem sequence is then classified as a direct retrieval type. Otherwise, it is classified as a summary report. .

[0054] In this step, through the collaborative modeling of semantic and structured features, the originally fuzzy classification decision is transformed into a stable judgment mechanism based on data-driven approaches, thereby achieving quantifiable and verifiable classification of the problem intent.

[0055] 2. Structured feature extraction and dimensionality reduction.

[0056] For government data tables within government data, PandasAI can be used to extract the table headers and a predetermined amount of standardized data (e.g., 10%) as target feature data, rather than the entire dataset. This ensures that the data scope is strictly limited to the header information and a limited number of standardized samples, thereby achieving minimal feature extraction from government data tables. The specific expression is as follows:

[0057]

[0058] In the formula, For the file header; For government data tables, among which For the number of records, Number of fields; Standardized data representing 10% of the total data volume; This is the sampling function.

[0059] 3. Generate dynamic query statements.

[0060] Construct a feature context table based on the target feature data. Then, the lightweight model is invoked to generate SQL query statements. .

[0061] 4. Multiple rounds of iterative verification and error correction.

[0062] Establish a verification function Determine the query statement The effectiveness.

[0063] like =1, then the current round of data analysis ends; if =0, then an error message will be sent. Feedback is fed back to the model for iterative correction until the correct SQL query is output or the iteration termination condition is met.

[0064] 5. Dynamic output generation.

[0065] Optionally, the government affairs analysis results corresponding to the government affairs data are determined based on the query results and the classification results of the question sequence, including: if the classification results corresponding to the question sequence are of the direct retrieval type, the query results are determined as the government affairs analysis results corresponding to the government affairs data; if the classification results corresponding to the question sequence are of the summary report type, the processing logic template corresponding to the question sequence is determined, and the question sequence and processing logic template are analyzed through a data processing model to obtain a summary report corresponding to the question sequence, and the summary report is determined as the government affairs analysis results corresponding to the government affairs data.

[0066] For example, suppose a query statement The corresponding query result is structured data R. Based on the classification results in the first step. Define output function ,in This is the final output result.

[0067] like If so, the query results output by the lightweight model will be returned directly; if Then, the heavyweight model (i.e., the data processing model mentioned above) is invoked to generate a summary report containing trend analysis, key indicators, and conclusions and recommendations based on the instruction constraint logic template.

[0068] In the above process, when government data includes problem sequences and government data tables, the problem sequences are classified to identify whether they belong to direct retrieval type or summary report type. Only the file headers and a small amount of standardized data (10% of the original data volume) in the government data tables are extracted as target feature data. This achieves lightweight data interaction, compressing the traditional 15-minute data analysis task to within 3 minutes, improving the response speed by 80%. At the same time, compared with the traditional full data transmission solution, the risk of data privacy leakage is reduced by 90%.

[0069] Furthermore, if the classification result is a direct retrieval type, the query result returned by the lightweight model is directly output as the government affairs analysis result, achieving a rapid response. If the classification result is a summary report type, a preset processing logic template corresponding to the semantics of the question sequence is further matched, and the data processing model integrates the question sequence intent and query results, generating a standardized summary report containing trend analysis, key indicators, and conclusions and recommendations based on the template in a structured manner. This avoids the resource abuse of large models, reduces inference costs by 56%, and effectively solves the problems of inconsistent analysis result formats and incomplete content under complex analysis needs. It realizes the automated and standardized transformation from raw data to compliant government affairs analysis reports, improving the efficiency, accuracy, and compliance of government affairs data processing.

[0070] Figure 4 This is a data processing flowchart of an intelligent government affairs system in an intelligent meeting drafting scenario according to an embodiment of this application. The data processing logic is as follows: When the government affairs data includes meeting data corresponding to the government affairs meeting drafting scenario, the meeting data is matched with a local meeting drafting knowledge base to obtain a matching result; if the matching result indicates that the meeting data and the meeting drafting knowledge base do not match, the meeting drafting result corresponding to the meeting data is directly output through a preset interaction model; if the matching result indicates that the meeting data and the meeting drafting knowledge base match, the reference document with the highest similarity to the meeting data is determined from the meeting drafting knowledge base, and the meeting drafting result corresponding to the meeting data is determined based on the reference document. The specific process analysis is as follows:

[0071] 1. Matching local meeting drafting knowledge base.

[0072] In intelligent government systems, a structured knowledge base for drafting meeting documents, built into historical document templates, serves as the core resource pool, which can be represented as follows: ,in, This indicates the meeting or draft document type in the i-th document template. This represents the meeting or drafting tag in the i-th document template, and N represents the total number of templates in the meeting drafting knowledge base.

[0073] In the context of drafting documents for government meetings, the meeting data contained within government data can first be defined as a set of four-tuples. ,in For meeting or document drafting purposes, For meeting or document drafting tags, For meeting or document drafting topics, Describe the specific requirements for the meeting or document drafting. Then, The two factors are uniquely matched against the local meeting drafting knowledge base. If no match is found, the set of quadruples is removed. Input a preset interaction model, and the system will directly output the meeting drafting results corresponding to the meeting data through the large model; if the match is successful, the semantic vector retrieval process will be triggered.

[0074] 2. Semantic vector retrieval and document template matching.

[0075] Optionally, the reference document with the highest similarity to the meeting data is determined from the meeting document knowledge base, including: generating a second semantic vector corresponding to the meeting data through a semantic vector model; determining a first similarity between the second semantic vector and a third semantic vector in the meeting document knowledge base, wherein the third semantic vector is the semantic vector corresponding to any document template in the meeting document knowledge base; determining a first similarity set, and determining the document template corresponding to the highest similarity in the first similarity set as the reference document corresponding to the meeting data, wherein the first similarity set is the similarity set with a first similarity greater than a first preset screening threshold.

[0076] For example, firstly, a set of quadruplets is generated using the BGE-M3 model. The corresponding topic vector (i.e., the second semantic vector mentioned above) Subsequently, the second semantic vector is calculated and compared with the local meeting drafting knowledge base vector pool. The Middle The cosine similarity between the semantic vectors (i.e., the third semantic vector mentioned above) is used to select the document template with the highest similarity from the first similarity set that is greater than a first preset screening threshold (e.g., 0.75) as the reference document. The specific expression is as follows:

[0077]

[0078]

[0079] In the formula, Indicates the first similarity. Represents the second semantic vector. Let M represent the third semantic vector, and M represent the total number of document templates in the meeting drafting knowledge base.

[0080] In this step, after the meeting data is successfully matched with the meeting document knowledge base, the BGE-M3 model generates a semantic vector corresponding to the meeting data and calculates the similarity between the semantic vector and the semantic vector corresponding to each document template in the knowledge base. Only the similarity results greater than 0.75 are retained to form a similarity set, and then the document template corresponding to the highest similarity is selected as the reference document. This effectively filters out low-relevance templates, avoids the risk of misselection due to an overly broad similarity evaluation range, and solves the technical problem of unstable document template recommendation quality due to the lack of a similarity screening mechanism.

[0081] 3. Large-scale model drives document generation.

[0082] Based on quadruples and the selected reference documents The DeepSeek-70B large model is used to iteratively generate meeting organization or draft documents and output the meeting organization draft documents.

[0083] 4. The knowledge base for drafting conference documents is dynamically updated.

[0084] After the intelligent government affairs system outputs the meeting draft results, users can modify the output content through a visual interface, generate revised content, and update the meeting draft knowledge base, including: receiving the first revised content of the meeting draft results detected through the visual interface; and updating the meeting draft knowledge base based on the first revised content.

[0085] Specifically, the system receives the first revisions made by users to the generated meeting drafts through a visual interface. These revisions are then used as feedback in real-world scenarios to directly update the corresponding knowledge entries in the local meeting draft knowledge base. This allows the knowledge base to dynamically evolve based on optimization behaviors in actual business operations, overcoming the problems of the original knowledge base content being fixed and unable to adapt to changes in usage habits and standards. This enables continuous optimization and closed-loop evolution of the knowledge base.

[0086] In the above process, when government data includes meeting data corresponding to the scenario of drafting government meeting documents, the meeting data is semantically matched with the locally deployed meeting document knowledge base to determine whether it has a high similarity to existing official document templates. If the matching result shows that the meeting data is highly similar to the reference official documents in the knowledge base, the meeting document is determined based on the existing official document with the highest similarity, which significantly reduces the risk of token consumption and privacy leakage. If the matching result shows that the meeting data is a new scenario and there are no similar reference official documents, a preset interaction model is used for intelligent generation to ensure flexibility and adaptability.

[0087] This "match first, generate later" dual-path mechanism effectively avoids redundant generation of existing templates. Compared with the traditional manual retrieval method, the material generation time is reduced from 30 minutes to 2 minutes. At the same time, it effectively improves the consistency and efficiency of official document output. The error rate of official document format is reduced from 12% in the traditional solution to 0.8%, realizing the localization, efficiency and security of the drafting process of government affairs meetings.

[0088] Figure 5This is a data processing flowchart of an intelligent government affairs system in a document distribution scenario according to an embodiment of this application. The data processing logic is as follows: When the government affairs data includes document data corresponding to the government affairs document distribution scenario, the document data is processed through a preset document decision model to obtain a distribution decision corresponding to the document data. The distribution decision includes at least the initial document distribution unit, initial document distribution content, and confidence level corresponding to the document data. If the confidence level is greater than or equal to a preset confidence threshold, the initial document distribution unit and initial document distribution content are determined as the document distribution result corresponding to the document data. If the confidence level is less than the preset confidence threshold, the document distribution template with the highest similarity to the document data is determined from the local document distribution knowledge base, and the target document distribution unit and target document distribution content in the document distribution template are determined as the document distribution result corresponding to the document data. The specific process analysis is as follows:

[0089] 1. Extraction of official document texts and rule-based weighting.

[0090] In the intelligent government affairs system, a structured document distribution knowledge base is also built-in based on historical document distribution templates, which can be represented as... ,in, This represents the receiving unit in the i-th document distribution template. This represents the title of the incoming document in the i-th document distribution template. Let N represent the main text content of the i-th document distribution template, and N represent the total number of templates in the document distribution knowledge base.

[0091] In the context of government document distribution, for the document data contained in government data, the first step is to extract the text and construct a structured feature set. ,in, Indicates the sending organization. This is the title of the article. This is the main text content. Subsequently, a document feature weight allocation function is constructed. ,in This indicates a feature concatenation operation. Next, the document features are... Document feature weight allocation function The input is processed using a pre-defined document decision-making model to generate a preliminary distribution decision. ,in As the initial document distribution unit, This is the initial document distribution content. This is the confidence score. Finally, a judgment is made based on a preset confidence threshold (e.g., 90). Then output directly and ,like Then proceed to the document distribution knowledge base retrieval process.

[0092] 2. Semantic vector retrieval and document distribution template matching.

[0093] Optionally, the document distribution template with the highest similarity to the document data is determined from the local document distribution knowledge base, including: generating a fourth semantic vector corresponding to the document data through a semantic vector model; determining a second similarity between the fourth semantic vector and a fifth semantic vector in the document distribution database, wherein the fifth semantic vector is the semantic vector corresponding to any document distribution template in the document distribution knowledge base; determining a second similarity set, and determining the document distribution template corresponding to the maximum similarity in the second similarity set as the document distribution template corresponding to the document data, wherein the second similarity set is a similarity set with a second similarity greater than a second preset screening threshold.

[0094] For example, firstly, the BGE-M3 model is used to generate document features. The corresponding topic vector (i.e., the fourth semantic vector mentioned above) , ,in , These represent the semantic vectors corresponding to the receiving unit and the title, respectively. Then, the second similarity or comprehensive confidence score between the fourth semantic vector and the i-th semantic vector in the local document distribution knowledge base (i.e., the fifth semantic vector mentioned above) is calculated. From the second similarity set that is greater than a second preset screening threshold (e.g., 0.9), the document distribution template with the highest similarity is selected. The specific expression is as follows:

[0095]

[0096] In the formula, Indicates the second similarity. , These represent the semantic vectors corresponding to the source unit and the title in the fourth semantic vector, respectively. , These represent the semantic vectors corresponding to the source unit and the title in the fifth semantic vector, respectively.

[0097] If it exists Then the document will be distributed to the corresponding knowledge base. and As the unit distributing official documents and the content of the document distribution, otherwise, the following selection decision will be made based on merit:

[0098]

[0099] In the formula, This indicates the final result of the document distribution.

[0100] In this step, when the confidence level output by the document decision model is lower than the preset confidence threshold, the BGE-M3 model generates a semantic vector corresponding to the document data and calculates the similarity between the semantic vector and the semantic vector corresponding to each document distribution template in the knowledge base. Only the high relevance similarity greater than 0.9 is retained. Then, the document distribution template corresponding to the one with the highest similarity from the filtered similarity set is selected as the final distribution basis. Thus, without relying on the full data upload and the long context processing of the large model, accurate semantic retrieval and intelligent reuse of historical distribution experience are achieved. This effectively avoids distribution errors caused by mismatching low relevance templates and improves the decision accuracy and system efficiency in low confidence scenarios.

[0101] 3. Update the document distribution knowledge base.

[0102] After the intelligent government affairs system outputs the document distribution results, users can modify the output content through a visual interface, generate revised content, and update the document distribution knowledge base, including: receiving the second revised content for the document distribution results detected through the visual interface; and updating the document distribution knowledge base based on the second revised content.

[0103] Specifically, the system receives the user's second revision of the distribution unit or distribution content through a visual interface, and dynamically updates the distribution unit and distribution content of the corresponding template in the local document distribution knowledge base based on the revision. This allows the knowledge base to continuously absorb the best practices of manual correction in actual business scenarios, thereby breaking through the limitations of the initial template data quality and realizing the automatic evolution and accurate adaptation of the knowledge base content as it is used, effectively improving the template reuse rate and the accuracy of distribution decisions.

[0104] In the above process, when government data includes document data corresponding to the document distribution scenario, the content is semantically analyzed and structured through a preset document decision model to generate a distribution decision containing the initial document distribution unit, initial document distribution content, and confidence level. A dual-path decision mechanism is implemented based on a preset confidence threshold: when the confidence level reaches or exceeds the threshold, the model output is directly used as the final distribution result to ensure high efficiency and automation; when the confidence level is below the threshold, the similarity retrieval of the local document distribution knowledge base is automatically triggered to accurately match historical distribution templates, and the model output is replaced by the verified target document distribution unit and target document distribution content. This allows for dynamic optimization by integrating local business experience when sufficient confidence is lacking, achieving multi-dimensional weighted retrieval of document distribution units and content. The error rate of distribution units and content has been reduced from 18% to 2%, effectively overcoming the problems of high distribution error rate and inability to adapt to regional or contextual distribution conventions caused by the reliance on a single model judgment in traditional methods. Meanwhile, the document distribution knowledge base is dynamically updated based on user feedback to adapt to personalized government office scenarios. The knowledge base reuse rate has increased from the initial 35% to 82%, effectively improving the accuracy of document distribution and enhancing the system's adaptability to complex government scenarios.

[0105] In this application embodiment, a three-in-one technical architecture of "lightweight data interaction + hybrid vector semantic retrieval + dual-threshold dynamic decision-making with confidence level" is constructed for the first time in an intelligent government affairs system. For intelligent analysis scenarios such as ultra-large table data, a novel approach is adopted: head semantic parsing + 10% standardized sampling replaces the full data upload. Combined with an intelligent classifier based on question type (direct retrieval / summary report), the analysis efficiency is effectively improved. For scenarios such as document generation and distribution, a novel approach is adopted: BGE-M3 multi-vector semantic modeling and dynamic threshold matching mechanism are integrated. Through a dual-threshold optimization decision-making model of AI confidence level + knowledge base similarity, the self-evolutionary update of the knowledge base is achieved while ensuring the standardization of generation. This truly solves the core pain points of traditional intelligent government affairs systems, such as reluctance to transmit data, inability to use templates, and difficulty in accumulating experience. It creates a new paradigm of government affairs AI with low resource consumption, high security and reliability, and strong adaptability.

[0106] Overall, this application can collaboratively support semantic vector retrieval and dynamic knowledge base update mechanisms for multiple scenarios such as intelligent analysis, intelligent meeting drafting, and document distribution. This enables the system to complete high-precision document template matching and distribution decisions without relying on the full data upload, achieving a synergistic optimization effect of improved response speed, reduced error rate, and increased knowledge reuse rate.

[0107] According to embodiments of this application, a government data processing apparatus is provided. It should be noted that the government data processing apparatus of this application can be used to execute the government data processing method provided in the embodiments of this application. The government data processing apparatus provided in the embodiments of this application will be described below.

[0108] Figure 6 This is a structural diagram of a government data processing device according to an embodiment of this application. Figure 6 As shown, the device includes:

[0109] The receiving module 60 is used to receive government data input by the user;

[0110] The first determining module 62 is used to determine the classification result corresponding to the problem sequence and extract the target feature data from the government data table when the government data contains a problem sequence and a government data table corresponding to the government analysis scenario. The classification result is used to indicate the problem type of the problem sequence, and the target feature data only contains the file header of the government data table and standardized data of a preset amount.

[0111] The second determining module 64 is used to determine the query results corresponding to the target feature data, and to determine the government affairs analysis results corresponding to the government affairs data based on the query results and the classification results of the question sequence.

[0112] By using the receiving module, the first determining module, and the second determining module in the aforementioned government data processing device, the goal of efficient government data processing through lightweight feature extraction and intelligent classification is achieved. This reduces the resource consumption of large models and the risk of privacy leakage, thereby solving the technical problem of poor data processing efficiency in government scenarios caused by resource and privacy risks in data processing of intelligent government systems in related technologies, and the inability to effectively process table data exceeding the token length limit.

[0113] In the government data processing device provided in this application embodiment, the first determining module is further configured to generate a first semantic vector corresponding to the question sequence through a semantic vector model; extract structured features from the question sequence, wherein the structured features include at least the question length and keyword frequency of the question sequence; determine a joint semantic vector corresponding to the question sequence based on the first semantic vector and the structured features; process the joint semantic vector through a classifier and output a probability value corresponding to the question sequence, wherein the probability value is used to indicate the probability that the question sequence belongs to different question types; when the probability value is greater than or equal to a preset probability threshold, determine the classification result corresponding to the question sequence as a direct retrieval type, and when the probability value is less than the preset probability threshold, determine the classification result corresponding to the question sequence as a summary report type.

[0114] In the government data processing device provided in this application embodiment, the second determining module is further configured to determine the query result as the government analysis result corresponding to the government data when the classification result corresponding to the problem sequence is of the direct retrieval type; and to determine the processing logic template corresponding to the problem sequence when the classification result corresponding to the problem sequence is of the summary report type, and to analyze the problem sequence and the processing logic template through the data processing model to obtain the summary report corresponding to the problem sequence, and to determine the summary report as the government analysis result corresponding to the government data.

[0115] The government data processing device provided in this application embodiment also includes a matching module 66, which is used to match the meeting data with a local meeting drafting knowledge base when the government data contains meeting data corresponding to the meeting drafting scenario, and obtain a matching result; when the matching result indicates that the meeting data does not match the meeting drafting knowledge base, the module directly outputs the meeting drafting result corresponding to the meeting data through a preset interaction model; when the matching result indicates that the meeting data matches the meeting drafting knowledge base, the module determines the reference document with the highest similarity to the meeting data from the meeting drafting knowledge base, and determines the meeting drafting result corresponding to the meeting data based on the reference document.

[0116] In the government data processing device provided in this application embodiment, the matching module is further configured to generate a second semantic vector corresponding to the meeting data through a semantic vector model; determine a first similarity between the second semantic vector and a third semantic vector in the meeting document knowledge base, wherein the third semantic vector is a semantic vector corresponding to any document template in the meeting document knowledge base; determine a first similarity set, and determine the document template corresponding to the maximum similarity in the first similarity set as a reference document corresponding to the meeting data, wherein the first similarity set is a similarity set with a first similarity greater than a first preset screening threshold.

[0117] In the government data processing device provided in this application embodiment, the matching module is further configured to receive the first revised content of the meeting drafting results detected through the visual interface; and update the meeting drafting knowledge base according to the first revised content.

[0118] The government data processing device provided in this application embodiment further includes a processing module 68, which is used to process the government data through a preset document decision model when the government data contains document data corresponding to the government document distribution scenario, to obtain a distribution decision corresponding to the document data. The distribution decision includes at least the initial document distribution unit, the initial document distribution content, and the confidence level corresponding to the document data. If the confidence level is greater than or equal to a preset confidence threshold, the initial document distribution unit and the initial document distribution content are determined as the document distribution result corresponding to the document data. If the confidence level is less than the preset confidence threshold, the document distribution template with the highest similarity to the document data is determined from the local document distribution knowledge base, and the target document distribution unit and the target document distribution content in the document distribution template are determined as the document distribution result corresponding to the document data.

[0119] In the government data processing device provided in this application embodiment, the processing module is further configured to generate a fourth semantic vector corresponding to the official document data through a semantic vector model; determine a second similarity between the fourth semantic vector and a fifth semantic vector in the official document distribution database, wherein the fifth semantic vector is a semantic vector corresponding to any official document distribution template in the official document distribution knowledge base; determine a second similarity set, and determine the official document distribution template corresponding to the maximum similarity in the second similarity set as the official document distribution template corresponding to the official document data, wherein the second similarity set is a similarity set in which the second similarity is greater than a second preset screening threshold.

[0120] In the government data processing device provided in this application embodiment, the processing module is further configured to receive the second revision content of the document distribution result detected through the visual interface; and update the document distribution knowledge base according to the second revision content.

[0121] This application also provides an electronic device, including: a memory and a processor, wherein the memory is used to store program instructions; and the processor is connected to the memory and used to execute the above-described government data processing method.

[0122] It should be noted that the aforementioned electronic equipment is used to perform Figure 2 The government data processing method shown above also applies to this electronic device, and will not be repeated here.

[0123] This application also provides a non-volatile storage medium, which includes a stored computer program, wherein the device containing the non-volatile storage medium executes the above-mentioned government data processing method by running the computer program.

[0124] It should be noted that the aforementioned non-volatile storage media is used for execution. Figure 2The government data processing method shown above also applies to this non-volatile storage medium, and will not be repeated here.

[0125] This application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the above-described government data processing method.

[0126] It should be noted that the above-mentioned computer program product is used to execute Figure 2 The government data processing method shown above is also applicable to this computer program product, and will not be repeated here.

[0127] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0128] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0129] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0130] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0131] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0132] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0133] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for processing government data, characterized in that, include: Receive government data input by users; When the government data includes a problem sequence and a government data table corresponding to the government analysis scenario, the classification result corresponding to the problem sequence is determined, and the target feature data in the government data table is extracted. The classification result is used to indicate the problem type of the problem sequence, and the target feature data only includes the file header and standardized data of a preset amount in the government data table. Determine the query results corresponding to the target feature data, and determine the government affairs analysis results corresponding to the government affairs data based on the query results and the classification results of the question sequence.

2. The method according to claim 1, characterized in that, Determining the classification result corresponding to the question sequence includes: A first semantic vector corresponding to the question sequence is generated using a semantic vector model; Extract structured features from the question sequence, wherein the structured features include at least the question length and keyword frequency of the question sequence; Based on the first semantic vector and the structured features, a joint semantic vector corresponding to the question sequence is determined; The joint semantic vector is processed by a classifier to output a probability value corresponding to the question sequence, wherein the probability value is used to indicate the probability that the question sequence belongs to different question types; If the probability value is greater than or equal to a preset probability threshold, the classification result corresponding to the question sequence is determined to be a direct retrieval type; if the probability value is less than the preset probability threshold, the classification result corresponding to the question sequence is determined to be a summary report type.

3. The method according to claim 2, characterized in that, Based on the query results and the classification results of the question sequence, the government affairs analysis results corresponding to the government affairs data are determined, including: If the classification result corresponding to the question sequence belongs to the direct retrieval type, the query result will be determined as the government affairs analysis result corresponding to the government affairs data; If the classification result corresponding to the problem sequence belongs to the summary report type, a processing logic template corresponding to the problem sequence is determined, and the problem sequence and the processing logic template are analyzed through a data processing model to obtain a summary report corresponding to the problem sequence. The summary report is then identified as the government affairs analysis result corresponding to the government affairs data.

4. The method according to claim 1, characterized in that, The method further includes: If the government data includes meeting data corresponding to the scenario of drafting government meeting documents, the meeting data is matched with the local meeting document knowledge base to obtain the matching result; If the matching result indicates that the meeting data does not match the meeting drafting knowledge base, the meeting drafting result corresponding to the meeting data is directly output through a preset interaction model. If the matching result indicates that the meeting data matches the meeting drafting knowledge base, the reference document with the highest similarity to the meeting data is determined from the meeting drafting knowledge base, and the meeting drafting result corresponding to the meeting data is determined based on the reference document.

5. The method according to claim 4, characterized in that, The reference documents with the highest similarity to the meeting data are identified from the meeting drafting knowledge base, including: A second semantic vector corresponding to the meeting data is generated using a semantic vector model; Determine the first similarity between the second semantic vector and the third semantic vector in the meeting drafting knowledge base, wherein the third semantic vector is the semantic vector corresponding to any document template in the meeting drafting knowledge base; A first similarity set is determined, and the document template corresponding to the maximum similarity in the first similarity set is determined as the reference document corresponding to the meeting data, wherein the first similarity set is the set of similarities where the first similarity is greater than a first preset screening threshold.

6. The method according to claim 4, characterized in that, The method further includes: Receive the first revision of the meeting drafting results detected through the visual interface; Update the meeting draft knowledge base according to the first revision.

7. The method according to claim 1, characterized in that, The method further includes: When the government data includes document data corresponding to the government document distribution scenario, the document data is processed by a preset document decision model to obtain a distribution decision corresponding to the document data. The distribution decision includes at least the initial document distribution unit, the initial document distribution content, and the confidence level corresponding to the document data. If the confidence level is greater than or equal to a preset confidence threshold, the initial document distribution unit and the initial document distribution content are determined as the document distribution result corresponding to the document data; If the confidence level is less than the preset confidence threshold, the document distribution template with the highest similarity to the document data is determined from the local document distribution knowledge base, and the target document distribution unit and target document distribution content in the document distribution template are determined as the document distribution result corresponding to the document data.

8. The method according to claim 7, characterized in that, The document distribution template with the highest similarity to the document data is determined from the local document distribution knowledge base, including: A fourth semantic vector corresponding to the document data is generated using a semantic vector model; Determine the second similarity between the fourth semantic vector and the fifth semantic vector in the document distribution database, wherein the fifth semantic vector is the semantic vector corresponding to any document distribution template in the document distribution knowledge base; A second similarity set is determined, and the document distribution template corresponding to the maximum similarity in the second similarity set is determined as the document distribution template corresponding to the document data. The second similarity set is the set of similarities where the second similarity is greater than the second preset screening threshold.

9. The method according to claim 8, characterized in that, The method further includes: Receive the second revision of the document distribution result detected through the visual interface; Update the document distribution knowledge base according to the second revision.

10. A government data processing device, characterized in that, include: The receiving module is used to receive government data input by the user; The first determining module is used to determine the classification result corresponding to the problem sequence and extract the target feature data from the government data table when the government data contains a problem sequence and a government data table corresponding to the government analysis scenario. The classification result is used to indicate the problem type of the problem sequence, and the target feature data only contains the file header and standardized data of a preset amount from the government data table. The second determining module is used to determine the query results corresponding to the target feature data, and to determine the government affairs analysis results corresponding to the government affairs data based on the query results and the classification results of the question sequence.

11. An electronic device, characterized in that, include: A memory and a processor, wherein the memory is used to store program instructions; The processor, connected to the memory, is used to execute the government data processing method according to any one of claims 1 to 9.

12. A non-volatile storage medium, characterized in that, The non-volatile storage medium includes a stored computer program, wherein the device containing the non-volatile storage medium executes the government data processing method according to any one of claims 1 to 9 by running the computer program.

13. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the government data processing method according to any one of claims 1 to 9.

Images