Rural-level affair intelligent auxiliary system and method based on artificial intelligence

Abstract

The invention discloses a village-level affair intelligent auxiliary system and method based on artificial intelligence, and the method comprises the following steps: constructing and loading an HSTF model, collecting a multi-source input information set, and carrying out the semantic fusion processing; extracting entity information and a relation mapping set, and establishing a village-level transaction knowledge graph; performing dialect voice transcription and semantic partitioning based on the conference voice to generate a structured conference record; analyzing the report template set to extract field constraint information, executing field matching, data filling and consistency verification, and generating a standardized report file set; a document writing demand is received, and a standardized document first draft is generated in combination with the standardized report file set; receiving a village query request and a file uploading task, and generating a query result and a file version record; and adjusting model parameters and updating the knowledge graph based on the task feedback information. According to the invention, the intelligentization and processing precision of village management are improved.

Description

Technical Field

[0001] This invention relates to the field of e-government, and in particular to an intelligent auxiliary system and method for village-level affairs based on artificial intelligence. Background Technology

[0002] In the current administrative village affairs management work, meeting minutes, report filling, document drafting, village information inquiry, and document management still largely rely on manual work, resulting in problems such as low efficiency, poor accuracy, scattered data, and inconsistent formats. Village cadres often use handwriting or audio recording to record the content of meetings, and compiling the minutes takes several hours, with serious information omissions and non-standard formats, affecting the execution of subsequent tasks and the tracing of historical data.

[0003] During the reporting and data submission process, village-level staff must manually search for and input data from scattered ledgers, forms, and electronic files, which easily leads to input errors and omissions. Asynchronous data updates result in distorted statistical results, impacting decision-making and analysis by township and higher-level departments. Furthermore, the document drafting process lacks standardized templates and policy-related mechanisms, often relying on experience for editing, resulting in lengthy processes, significant formatting inconsistencies, and non-compliance with government regulations.

[0004] Meanwhile, village information and policy documents are scattered across paper archives, Excel files, and personal computers, resulting in low retrieval efficiency, untimely document updates, and inconsistent versioning, making it difficult to quickly obtain accurate information. Most existing information systems only possess basic document storage and retrieval functions, failing to achieve data linkage and intelligent assistance between village-level affairs such as meetings, reports, documents, and inquiries.

[0005] Therefore, how to provide an intelligent assistance system and method for village affairs based on artificial intelligence is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose an intelligent auxiliary system and method for village-level affairs based on artificial intelligence. This invention comprehensively utilizes semantic hierarchical coding, context alignment, knowledge graph construction, and self-supervised learning techniques to achieve functions such as meeting record generation, automatic report filling, intelligent document writing, document version management, and task feedback optimization in village-level affairs. By constructing an HSTF model, it achieves semantic fusion and dynamic reasoning of multi-source information, forming a village-level affairs knowledge graph, and realizing a closed loop of dialect speech recognition, semantic understanding, and intelligent decision-making. This invention has advantages such as high automation, strong information processing consistency, high efficiency in transaction linkage, and adaptive model optimization.

[0007] An intelligent assistance method for village-level affairs based on artificial intelligence according to an embodiment of the present invention includes the following steps:

[0008] Build and load the HSTF model, collect multi-source input information sets, perform semantic fusion processing, extract report template sets, and generate a unified semantic representation structure and semantic confidence matrix;

[0009] Based on a unified semantic representation structure and semantic confidence matrix, a village-level affairs knowledge graph is established, along with a version identifier table and a node time index table.

[0010] Based on the collected meeting audio, dialect speech transcription and semantic segmentation are performed. Based on the village-level affairs knowledge graph, structured meeting minutes, a set of to-do items, and a set of responsible persons are generated.

[0011] Parse the report template set and extract field constraint information. Based on the village-level affairs knowledge graph, combine the to-do items set and the responsible person mapping set to perform field matching, data filling and field consistency verification to generate a standardized report file set.

[0012] Upon receiving document drafting requests, the system combines village-level affairs knowledge graphs, structured meeting minutes, and standardized report file sets to select template structures and element sets, generating standardized document drafts.

[0013] Receive village information query requests and file upload tasks, output query results and mark information source paths based on the semantic indexing mechanism of village-level affairs knowledge graph and HSTF model, and call the version identifier table and node time index table to generate file version records;

[0014] Based on the task execution feedback information, the model parameters are adjusted and the village-level affairs knowledge graph is updated to generate a set of intelligent assistance results for village-level affairs.

[0015] Optionally, the generation of the report template set, the unified semantic representation structure, and the semantic confidence matrix includes:

[0016] Collect meeting audio, document text, report templates, policy documents and basic data, perform format standardization and source marking, and aggregate to form a multi-source input information set;

[0017] Construct and load the HSTF model, which includes a semantic hierarchical coding unit, a context alignment unit, a speech recognition unit, a semantic reasoning unit, a semantic indexing mechanism, and a self-supervised reflow unit;

[0018] Load the parameter configurations of the semantic hierarchical coding unit and the context alignment unit to establish a processing channel for a multi-source input information set;

[0019] The semantic hierarchical coding unit is invoked to perform hierarchical semantic coding on the multi-source input information set to generate hierarchical semantic representation. The context alignment unit is invoked to perform cross-source context alignment and conflict resolution on the hierarchical semantic representation and merge them to obtain a unified semantic representation structure.

[0020] Based on the HSTF model, structural parsing and field extraction are performed on report templates in a multi-source input information set to extract a set of report templates;

[0021] Based on the calculation results of the HSTF model in the semantic hierarchical coding unit and the context alignment unit, the semantic confidence matrix is ​​calculated.

[0022] Optionally, the generation of the village-level affairs knowledge graph, version identifier table, and node time index table includes:

[0023] Based on the aligned semantic boundaries in the unified semantic representation structure and the confidence threshold in the semantic confidence matrix, entity extraction rules and relation extraction rules are determined.

[0024] Based on the entity extraction rules, six types of entity nodes are generated by extracting villager entities, meeting entities, policy entities, document entities, task entities, and data field entities from the unified semantic representation structure.

[0025] Generate a set of relationship mappings by generating membership relationships, initiation relationships, reference relationships, dependency relationships, and source mapping relationships according to the relationship extraction rules;

[0026] In the village affairs knowledge graph, population data nodes, land data nodes and financial data nodes are constructed. Population data nodes are connected to villager entities through source mapping relationships, land data nodes are connected to task entities through dependency relationships, and financial data nodes are connected to document entities through reference relationships.

[0027] Establish a version identifier table and a node time index table. Write the version status of the six types of entity nodes, population data nodes, land data nodes and fiscal data nodes into the version identifier table, and write the creation time and update time of the nodes into the node time index table.

[0028] Optionally, the generation of the structured meeting minutes, the to-do list set, and the responsible person mapping set includes:

[0029] The collected conference audio is input into the speech recognition unit of the HSTF model to perform acoustic preprocessing and dialect feature extraction, generating a speech feature stream;

[0030] Based on the speech feature stream, dialect speech is transcribed to generate conference transcripts, and semantic segmentation is performed to generate a set of semantic segment fragments.

[0031] Based on the village-level affairs knowledge graph, the semantic block fragment set is parsed for topic parsing, speech parsing, resolution parsing and time constraint parsing to generate structured meeting minutes;

[0032] Based on structured meeting minutes, task descriptions, responsible parties, and completion deadlines are extracted to generate a set of to-do items and a set of responsible party mappings.

[0033] Optionally, the generation of the standardized report file set includes:

[0034] Parse the report template set and extract field constraint information to determine the field name, field type, value range and filling rules;

[0035] Read the population data nodes, land data nodes, and financial data nodes in the village-level affairs knowledge graph, and establish the matching relationship between fields and data node attributes based on field constraint information;

[0036] The scope of the reporting objects is determined based on the set of to-do items and the set of responsible persons, and the matching relationship is limited to the set of node attributes within the scope of the reporting objects;

[0037] Data is populated into the set of report templates according to the matching relationships to generate draft versions of a set of standardized report files;

[0038] Perform field consistency checks on the draft version. The checks include cross-field logical consistency, value range consistency, and source path consistency. Obtain the check results.

[0039] The verification results are recorded as report verification records and written into the village-level affairs knowledge graph. Based on the verification results, the draft version is finalized to generate a standardized report file set.

[0040] Optionally, the generation of the standardized document draft includes:

[0041] It receives document drafting requests and reads village-level affairs knowledge graphs, structured meeting minutes, and standardized report files as input.

[0042] The semantic reasoning unit of the HSTF model is invoked to determine the template structure and element set based on the input content;

[0043] Based on the template structure, establish a one-to-one correspondence between the main text content and data fields and policy nodes, data source nodes, topic nodes, speaking nodes, resolution nodes, and time constraints;

[0044] Generate a standardized draft document based on the set of elements and their corresponding relationships, fill in the main text and data fields, and generate the title, main text paragraphs, signature, and date;

[0045] Establish a document citation index, which points to policy nodes and data source nodes, and records the information source path.

[0046] Optionally, the generation of the file version record includes:

[0047] Receive village information query requests and file upload tasks, read the village-level affairs knowledge graph, version identifier table and node time index table, and call the semantic indexing mechanism of the HSTF model;

[0048] Based on the semantic indexing mechanism, semantic parsing and condition mapping are performed on village information query requests. Matching nodes and matching relationships are retrieved from the village-level affairs knowledge graph, and query results are generated and the information source path is marked in the query results.

[0049] Based on the version identifier table and the node time index table, the file upload task is deduplicated, field compared and version tracked. The matching results are classified into three states: new record, replaced record and retained record, and the corresponding version identifier and time index information are recorded.

[0050] Based on the matching results, a file version record is generated, which records the version identifier, creation time, update time, and information source path.

[0051] Optionally, the generation of the village-level affairs intelligent assistance result set includes:

[0052] Receive and aggregate task execution feedback information, which includes confirmation results of structured meeting minutes, confirmation results of standardized report files, confirmation results of report verification records, confirmation results of standardized document drafts, confirmation results of document citation indexes, external execution status of query results, and external execution status of document version records;

[0053] The task execution feedback information is input into the self-supervised backflow unit of the HSTF model, sorted and aligned by source label and time label, to complete the input preparation for parameter adjustment;

[0054] The self-supervised backflow unit is invoked to adjust the HSTF model parameters based on the sorted and aligned task execution feedback information, resulting in the adjusted HSTF model parameters;

[0055] Based on the task execution feedback information, the entity confidence and relation weight in the village affairs knowledge graph are updated, the update results are written back to the village affairs knowledge graph, and kept consistent with the version identifier table and node time index table.

[0056] Based on the updated village-level affairs knowledge graph, standardized report files, standardized document drafts, query results, and document version records, the self-supervised backflow unit of the HSTF model is invoked to perform result aggregation operations, generating a set of intelligent auxiliary results for village-level affairs that includes task completion status, data consistency evaluation, information traceability path, and model parameter update summary.

[0057] According to an embodiment of the present invention, a village-level affairs intelligent assistance system based on artificial intelligence includes:

[0058] The multi-source data acquisition module is used to acquire multi-source data, perform format normalization, and generate a set of multi-source input information.

[0059] The HSTF model module is used to build and load HSTF models, which include semantic hierarchical coding units, context alignment units, speech recognition units, semantic reasoning units, semantic indexing mechanisms, and self-supervised reflow units.

[0060] The village-level affairs knowledge graph construction module is used to extract entity information based on a unified semantic representation structure and semantic confidence matrix, establish multi-dimensional relationships, and construct a village-level affairs knowledge graph.

[0061] The meeting minutes generation module is used to perform dialect speech transcription and semantic segmentation through the speech recognition unit to generate structured meeting minutes, a set of to-do items, and a set of responsible persons mappings;

[0062] The report generation and verification module is used to perform field matching, data filling and consistency verification based on the village-level affairs knowledge graph and the responsible person mapping set to generate a standardized report file set.

[0063] The document generation module is used to call the semantic reasoning unit and combine knowledge graphs, meeting minutes and report files to generate standardized document drafts and document citation indexes.

[0064] The query and file management module is used to output query results based on semantic indexing, perform file upload deduplication and version tracking, and generate file version records.

[0065] The intelligent feedback and optimization module is used to gather task execution feedback information, call the self-supervised backflow unit to adjust model parameters and update the knowledge graph, and generate a set of intelligent assistance results for village-level affairs.

[0066] The beneficial effects of this invention are:

[0067] First, by constructing an HSTF model, this invention achieves deep integration and semantic unification of multi-source information in village affairs, enabling meeting audio, document text, report templates, policy documents, and basic data to be identified, parsed, and associated in the same semantic space. This solves the problems of scattered and inconsistent formats of existing village affairs data. Through dialect speech recognition and semantic segmentation technology, the structured generation of meeting content is achieved, significantly improving the accuracy and efficiency of meeting record organization and reducing the workload of manual transcription and editing.

[0068] Secondly, by using knowledge graph-based multidimensional data modeling and semantic reasoning mechanisms, automatic report filling and intelligent document generation are achieved, ensuring consistency in content, format, and data association among various village affairs documents, and significantly improving the standardization and accuracy of village affairs processing.

[0069] Furthermore, this invention introduces a self-supervised feedback mechanism and a dynamic parameter optimization process, enabling the system to continuously adjust model weights based on task execution feedback, thereby achieving self-learning and continuous optimization of intelligent auxiliary functions. This effectively improves the automation level and governance efficiency of village affairs, and promotes the intelligentization, standardization, and efficiency of village-level government administration. Attached Figure Description

[0070] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0071] Figure 1 This is an overall flowchart of an intelligent auxiliary system and method for village affairs based on artificial intelligence proposed in this invention;

[0072] Figure 2 This is a schematic diagram of the HSTF model structure proposed in this invention;

[0073] Figure 3 This is a schematic diagram of the intelligent auxiliary closed-loop process for village-level affairs proposed in this invention. Detailed Implementation

[0074] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0075] refer to Figure 1-3 An intelligent assistance method for village-level affairs based on artificial intelligence includes the following steps:

[0076] Collect meeting audio, document text, report templates, policy documents and basic data to form a multi-source input information set, construct and load the HSTF model, use semantic hierarchical coding unit and context alignment unit to perform semantic fusion processing on the multi-source input information set, extract the report template set, and generate a unified semantic representation structure and semantic confidence matrix;

[0077] Based on a unified semantic representation structure and semantic confidence matrix, six types of entity information are extracted: villagers, meetings, policies, documents, tasks, and data fields. A set of relationship mappings is established, including membership, initiation, reference, dependency, and source mapping relationships. A village-level affairs knowledge graph is generated, and population data nodes, land data nodes, and fiscal data nodes are constructed in the village-level affairs knowledge graph to store structured data related to villagers' identities, land use, and fiscal revenue and expenditure. A version identifier table and a node time index table are also established to record the version status and time information of entity nodes.

[0078] The collected meeting audio is input into the speech recognition unit of the HSTF model, which performs dialect speech transcription and semantic segmentation. Based on the village-level affairs knowledge graph, structured meeting minutes are generated and the topic nodes, speaking nodes, resolution nodes and time constraints are written in. At the same time, a set of to-do items and a set of responsible persons are generated and updated to the village-level affairs knowledge graph.

[0079] The report template set is parsed and field constraint information is extracted. Based on the population data nodes, land data nodes and financial data nodes in the village-level affairs knowledge graph, and combined with the set of pending items and the set of responsible persons, field matching, data filling and field consistency verification are performed to generate a set of standardized report files. The report verification records are written into the village-level affairs knowledge graph.

[0080] Upon receiving document drafting requests, the semantic reasoning unit of the HSTF model is invoked. Combining the village-level affairs knowledge graph, structured meeting minutes, and standardized report file sets, a template structure and element set are selected to generate a standardized document draft. A document citation index is established to point to the policy nodes and data source nodes in the village-level affairs knowledge graph. The standardized document draft and document citation index are then written back to the village-level affairs knowledge graph.

[0081] Receive village information query requests and file upload tasks, output query results and mark information source paths based on the semantic indexing mechanism of village-level affairs knowledge graph and HSTF model, call the version identifier table and node time index table to perform deduplication, field comparison and version tracking on file upload tasks, generate file version records, and write the query results and file version records back to the village-level affairs knowledge graph to update the retrieval index and file status;

[0082] Collect task execution feedback information, including structured meeting minutes, standardized report files, report verification records, standardized document drafts, document citation indexes, query results, confirmation results of document version records, and external execution status. Call the self-supervised backflow unit of the HSTF model to adjust model parameters based on task execution feedback information and update entity confidence and relation weights in the village-level affairs knowledge graph to generate a set of intelligent auxiliary results for village-level affairs.

[0083] In this embodiment, the generation of the report template set, the unified semantic representation structure, and the semantic confidence matrix includes:

[0084] Collect meeting audio, document text, report templates, policy documents and basic data, perform format standardization and source marking, and aggregate to form a multi-source input information set;

[0085] Construct and load the HSTF model, which includes a semantic hierarchical coding unit, a context alignment unit, a speech recognition unit, a semantic reasoning unit, a semantic indexing mechanism, and a self-supervised reflow unit;

[0086] The semantic hierarchical coding unit performs hierarchical semantic coding on a multi-source input information set and jointly generates a unified semantic representation structure and semantic confidence matrix with the context alignment unit, while providing input for the construction of the village-level affairs knowledge graph; the context alignment unit performs cross-source semantic alignment and conflict resolution on the hierarchical semantic representation and outputs a unified semantic representation structure to support the extraction of report template sets and field constraint parsing; the speech recognition unit performs dialect speech transcription and semantic segmentation on meeting speech and outputs text fragments for the generation of structured meeting minutes and the extraction of to-do items; the semantic reasoning unit selects template structures and element sets based on the village-level affairs knowledge graph, structured meeting minutes, and standardized report file sets, and generates standardized document drafts and document citation indexes; the semantic indexing mechanism provides semantic retrieval and source path annotation for village information query and file upload tasks, and outputs semantic index information of query results and file version records; the self-supervised feedback unit adjusts model parameters and updates the weight configuration in the village-level affairs knowledge graph based on task execution feedback information to generate a set of intelligent auxiliary results for village-level affairs.

[0087] The HSTF model is specifically an improved Hierarchical Semantic Transform Fusion (HSTF) model.

[0088] Load the parameter configurations of the semantic hierarchical coding unit and the context alignment unit to establish a processing channel for a multi-source input information set;

[0089] The semantic hierarchical coding unit is invoked to perform hierarchical semantic coding on the multi-source input information set to generate hierarchical semantic representation. The context alignment unit is invoked to perform cross-source context alignment and conflict resolution on the hierarchical semantic representation and merge them to obtain a unified semantic representation structure.

[0090] Based on the HSTF model, structural parsing and field extraction are performed on report templates in a multi-source input information set to extract a set of report templates;

[0091] Based on the calculation results of the HSTF model in the semantic hierarchical coding unit and context alignment unit, the semantic confidence matrix is ​​calculated;

[0092] Output a unified semantic representation structure and semantic confidence matrix for use in building village-level transaction knowledge graphs, and output a set of report templates for generating standardized report files.

[0093] In this embodiment, the generation of the village-level affairs knowledge graph, version identifier table, and node time index table includes:

[0094] Receive a unified semantic representation structure and a semantic confidence matrix, and determine entity extraction rules and relation extraction rules based on the aligned semantic boundaries in the unified semantic representation structure and the confidence threshold in the semantic confidence matrix;

[0095] According to the entity extraction rules, villager entities, meeting entities, policy entities, document entities, task entities, and data field entities are extracted from the unified semantic representation structure, generating six types of entity nodes and assigning unique identifiers, which are then written into the village-level affairs knowledge graph.

[0096] According to the relationship extraction rules, the subordinate relationship, initiating relationship, referencing relationship, dependency relationship and source mapping relationship are generated to form a relationship mapping set. The relationship mapping set is written into the village-level transaction knowledge graph, and the relationship confidence is calculated for each relationship record by the semantic confidence matrix.

[0097] In the village-level affairs knowledge graph, population data nodes, land data nodes, and fiscal data nodes are constructed. Population data nodes are connected to villager entities through source mapping relationships, land data nodes are connected to task entities through dependency relationships, and fiscal data nodes are connected to document entities through reference relationships. The relationship confidence of nodes and relationships is recorded to store structured data related to villager identity, land use, and fiscal revenue and expenditure.

[0098] Establish a version identifier table and a node time index table. Write the version status of the six types of entity nodes, population data nodes, land data nodes and fiscal data nodes into the version identifier table. Write the creation time and update time of the nodes into the node time index table. Then associate the version identifier table and the node time index table with the village-level affairs knowledge graph.

[0099] The output includes a village-level affairs knowledge graph containing six types of entity nodes, a set of relationship mappings, population data nodes, land data nodes, and fiscal data nodes, for use in generating structured meeting minutes and standardized document drafts; the output also includes a version identifier table and a node time index table for use in generating query results and document version records.

[0100] In this embodiment, the generation of the structured meeting minutes, the to-do list set, and the responsible person mapping set includes:

[0101] The collected conference audio is input into the speech recognition unit of the HSTF model to perform acoustic preprocessing and dialect feature extraction, generating a speech feature stream;

[0102] Based on the speech feature stream, dialect speech transcription is performed to generate conference transcripts, and the semantic hierarchical coding unit is called to perform semantic segmentation on the conference transcripts to generate a set of semantic segment fragments;

[0103] Based on the village-level affairs knowledge graph, the semantic block fragment set is parsed for topic parsing, speech parsing, resolution parsing and time constraint parsing, generating structured meeting minutes, and writing the topic nodes, speech nodes, resolution nodes and time constraint relationships into the village-level affairs knowledge graph;

[0104] Based on the structured meeting minutes, task descriptions, responsible parties and completion deadlines are extracted to generate a set of to-do items and a set of responsible persons mappings, and the set of to-do items and the set of responsible persons mappings are written back to the village-level affairs knowledge graph;

[0105] Output structured meeting minutes for use in generating standardized document drafts, and output a set of to-do items and a set of responsible persons for use in generating a set of standardized report documents.

[0106] In this embodiment, the generation of the standardized report file set includes:

[0107] Parse the report template set and extract field constraint information to determine the field name, field type, value range and filling rules;

[0108] Read the population data nodes, land data nodes, and financial data nodes in the village-level affairs knowledge graph, and establish the matching relationship between fields and data node attributes based on field constraint information;

[0109] The scope of the reporting objects is determined based on the set of to-do items and the set of responsible persons, and the matching relationship is limited to the set of node attributes within the scope of the reporting objects;

[0110] Data is populated into the set of report templates according to the matching relationships to generate draft versions of a set of standardized report files;

[0111] Perform field consistency checks on the draft version. The checks include cross-field logical consistency, value range consistency, and source path consistency. Obtain the check results.

[0112] The verification results are recorded as report verification records and written into the village-level affairs knowledge graph. Based on the verification results, the draft version is finalized to generate a standardized report file set.

[0113] Output a set of standardized report files for use in generating standard document drafts, and output report verification records for use in collecting task execution feedback information.

[0114] In this embodiment, the generation of the standardized document draft includes:

[0115] It receives document drafting requests and reads village-level affairs knowledge graphs, structured meeting minutes, and standardized report files as input.

[0116] The semantic reasoning unit of the HSTF model is invoked to determine the template structure and element set based on the input content;

[0117] Based on the template structure, establish a one-to-one correspondence between the main text content and data fields and policy nodes, data source nodes, topic nodes, speaking nodes, resolution nodes, and time constraints;

[0118] Generate a standardized draft document based on the set of elements and their corresponding relationships, fill in the main text and data fields, and generate the title, main text paragraphs, signature, and date;

[0119] Establish a document citation index, which points to policy nodes and data source nodes, and records the information source path;

[0120] The standardized document draft and document citation index are written back to the village affairs knowledge graph. The standardized document draft is output for use in the intelligent auxiliary result collection of village affairs, and the document citation index is output for use in the information source path labeling.

[0121] In this embodiment, the generation of the file version record includes:

[0122] Receive village information query requests and file upload tasks, read the village-level affairs knowledge graph, version identifier table and node time index table, and call the semantic indexing mechanism of the HSTF model;

[0123] Based on the semantic indexing mechanism, semantic parsing and condition mapping are performed on village information query requests. Matching nodes and matching relationships are retrieved from the village-level affairs knowledge graph, and query results are generated and the information source path is marked in the query results.

[0124] Based on the version identifier table and the node time index table, the file upload task is deduplicated, field compared and version tracked. The matching results are classified into three states: new record, replaced record and retained record, and the corresponding version identifier and time index information are recorded.

[0125] Based on the matching results, a file version record is generated, which records the version identifier, creation time, update time, and information source path.

[0126] Write the query results and file version records back to the village-level affairs knowledge graph, update the retrieval index and file status, and output the query results and file version records for use in the aggregation of task execution feedback information.

[0127] In this embodiment, the generation of the village-level affairs intelligent assistance result set includes:

[0128] Receive and aggregate task execution feedback information, which includes confirmation results of structured meeting minutes, confirmation results of standardized report files, confirmation results of report verification records, confirmation results of standardized document drafts, confirmation results of document citation indexes, external execution status of query results, and external execution status of document version records;

[0129] The task execution feedback information is input into the self-supervised backflow unit of the HSTF model, sorted and aligned by source label and time label, to complete the input preparation for parameter adjustment;

[0130] The self-supervised backflow unit is invoked to adjust the HSTF model parameters based on the sorted and aligned task execution feedback information, resulting in the adjusted HSTF model parameters;

[0131] Based on the task execution feedback information, the entity confidence and relation weight in the village affairs knowledge graph are updated, the update results are written back to the village affairs knowledge graph, and kept consistent with the version identifier table and node time index table.

[0132] Based on the updated village affairs knowledge graph, standardized report files, standardized document drafts, query results, and document version records, the self-supervised backflow unit of the HSTF model is invoked to perform result aggregation operations. During the result aggregation process, the updated entity confidence and relation weights are used as weights to perform multi-dimensional fusion of structured meeting minutes, report verification records, document citation indexes, and external execution status. The fusion results are classified and integrated according to task category, responsible person, execution time limit, and data source path to generate a village affairs intelligent auxiliary result set that includes task completion status, data consistency evaluation, information traceability path, and model parameter update summary. The village affairs intelligent auxiliary result set is output for displaying the village affairs execution status and for continuous model optimization.

[0133] An artificial intelligence-based intelligent assistance system for village affairs includes:

[0134] The multi-source data acquisition module is used to collect conference audio, document text, report templates, policy documents and basic data, perform format standardization and source marking, and aggregate to form a multi-source input information set to provide input for subsequent semantic fusion and knowledge graph construction.

[0135] The HSTF model module is used to build and load the HSTF model, which includes a semantic hierarchical coding unit, a context alignment unit, a speech recognition unit, a semantic reasoning unit, a semantic indexing mechanism, and a self-supervised backflow unit. Specifically, the semantic hierarchical coding unit performs hierarchical semantic coding and generates a unified semantic representation structure; the context alignment unit performs cross-source semantic alignment and conflict resolution; the speech recognition unit performs dialect speech transcription and semantic segmentation; the semantic reasoning unit generates standardized document drafts and document citation indexes; the semantic indexing mechanism outputs query results and document version records; and the self-supervised backflow unit adjusts model parameters based on task execution feedback and generates a set of intelligent assistance results for village-level affairs.

[0136] The village-level affairs knowledge graph construction module is used to extract six types of entity information—villagers, meetings, policies, documents, tasks, and data fields—based on a unified semantic representation structure and semantic confidence matrix. It establishes affiliation, initiation, reference, dependency, and source mapping relationships to construct a village-level affairs knowledge graph. Within the graph, population data nodes, land data nodes, and financial data nodes are established. Additionally, a version identifier table and a node time index table are created to record node version status and time information.

[0137] The meeting minutes generation module is used to input the meeting audio into the speech recognition unit of the HSTF model to perform dialect speech transcription and semantic segmentation, combine it with the village-level affairs knowledge graph to generate structured meeting minutes, and write topic nodes, speaking nodes, resolution nodes and time constraint relationships, generate a set of to-do items and a set of responsible persons mapping and update them synchronously to the village-level affairs knowledge graph;

[0138] The report generation and verification module is used to parse the report template set and extract field constraint information. Based on the population data nodes, land data nodes and financial data nodes in the village-level affairs knowledge graph, and combined with the set of pending items and the set of responsible persons mapping, it performs field matching, data filling and consistency verification to generate a set of standardized report files and writes the report verification records into the village-level affairs knowledge graph.

[0139] The document generation module is used to receive document writing requests, call the semantic reasoning unit of the HSTF model, combine the village-level affairs knowledge graph, structured meeting minutes and standardized report file set to select template structure and element set, generate standardized document drafts, establish document citation index and write back to the village-level affairs knowledge graph;

[0140] The query and file management module is used to receive village information query requests and file upload tasks. Based on the semantic indexing mechanism of the HSTF model, it outputs query results and marks the information source path. It calls the version identifier table and node time index table to perform deduplication, field comparison and version tracking, generate file version records and update them to the village-level affairs knowledge graph.

[0141] The intelligent feedback and optimization module is used to collect task execution feedback information, including structured meeting minutes, standardized report files, report verification records, standardized document drafts, document citation indexes, confirmation results of query results and document version records, and external execution status. It calls the self-supervised backflow unit of the HSTF model to adjust model parameters and update entity confidence and relationship weights in the village affairs knowledge graph based on the feedback information. Based on the updated results, it performs result aggregation and multi-dimensional fusion to generate and output a set of intelligent auxiliary results for village affairs, which is used for displaying the village affairs execution status and continuously optimizing the model.

[0142] Example 1:

[0143] To verify the feasibility of this invention in practice, it was applied to a village-level affairs comprehensive management auxiliary system. An artificial intelligence model was used to uniformly process village affairs meetings, policy implementation, document drafting, report generation, and task tracking, thereby verifying its performance in semantic fusion, data consistency, and task feedback optimization. Experimental data was selected from the village affairs information platform, including 874 meeting audio recordings, 256 policy documents, 1120 written documents, 46 report templates, and approximately 250,000 pieces of basic financial, population, and land data.

[0144] In actual operation, the system first utilizes a multi-source data acquisition module to uniformly format and mark the sources of meeting audio, document text, policy documents, and report templates, and then inputs them into the HSTF model module. The semantic hierarchical encoding unit performs hierarchical encoding on information from different sources, while the context alignment unit achieves cross-source semantic unification. Subsequently, the system constructs a village-level affairs knowledge graph, extracting six types of entity nodes: villagers, meetings, policies, documents, tasks, and data fields, and establishing membership, initiation, dependency, and reference relationships. Based on this, the meeting record generation module automatically generates structured meeting records, the report generation and verification module completes data filling and consistency verification, the document generation module automatically generates standardized document drafts and establishes a reference index, the query and file management module performs file version comparison and index updates, and the intelligent feedback and optimization module continuously adjusts model parameters and entity confidence based on task feedback, achieving a closed-loop operation of intelligent assistance.

[0145] Table 1. Performance Comparison between the Invention System and the Traditional Village Affairs Management System

[0146] Indicator Categories Traditional system average value The average value of the system of this invention Increase Dialect speech recognition accuracy (%) 81.4 88.2 +8.8 Document generation time (minutes) 27 14 -48.1% Report field error rate (%) 6.5 3.1 -52.3% Data retrieval response time (seconds) 1.9 1.02 -46.3% Relationship mapping accuracy (%) 85.6 91.1 +5.5 Model adaptive update frequency / month No automatic updates 15 — System stability rate (%) 88.8 92.7 +4.4 Document citation index completeness (%) 85.2 92.4 +7.2 Semantic fusion similarity (correlation coefficient) 0.82 0.94 +0.12 Number of knowledge graph nodes (in ten thousand) 4.3 15.7 +265%

[0147] As shown in Table 1, the system of this invention significantly outperforms traditional systems in terms of dialect speech recognition, document generation efficiency, relation mapping accuracy, and semantic fusion capabilities. The dialect speech recognition accuracy is improved by 8.8 percentage points, demonstrating the strong robustness of the semantic hierarchical coding unit of the HSTF model in dialect feature processing and contextual association. Document generation time is reduced by 48.1%, indicating that the semantic reasoning unit can automatically match template structure and content elements, reducing manual editing time and improving consistency.

[0148] The 52.3% decrease in report field error rate indicates that the field constraint and responsible person mapping mechanism effectively improved the accuracy of table generation and validation. The 46.3% reduction in data retrieval response time reflects the high efficiency of the semantic indexing mechanism in retrieval path optimization and information association calculation. The 5.5 percentage point improvement in relationship mapping accuracy demonstrates that the village-level transaction knowledge graph possesses higher accuracy in semantic relationship recognition and dependency mapping among multiple entities.

[0149] The document citation index completeness improved by 7.2%, demonstrating that the system can accurately establish the correspondence between generated documents and policy documents and data source nodes, enhancing content traceability and logical verifiability. The system's stable operation rate improved by 4.4%, indicating that the self-supervised feedback unit can maintain the convergence and continuous reliability of model parameters in dynamic feedback optimization. The semantic fusion similarity improved to 0.94, indicating that the HSTF model achieves highly consistent semantic alignment across speech, text, and structured data.

[0150] The number of knowledge graph nodes has increased to 157,000, an increase of about 265% compared to traditional systems, indicating that the system has stronger multi-source information fusion and hierarchical expansion capabilities, providing rich structured support for the intelligent association and in-depth management of village-level affairs data.

[0151] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An artificial intelligence-based intelligent assistance method for village-level affairs, characterized in that, The method comprises the following steps: constructing and loading the HSTF model, collecting a multi-source input information set, performing semantic fusion processing, extracting a report template set, generating a unified semantic representation structure and a semantic confidence matrix; based on the unified semantic representation structure and the semantic confidence matrix, establishing a village-level transaction knowledge graph, and establishing a version identification table and a node time index table; based on the collected conference voice, performing dialect voice transcription and semantic segmentation, based on the village-level transaction knowledge graph, generating a structured conference record, a to-do list set and a responsibility person mapping set; parsing the report template set and extracting field constraint information, performing field matching, data filling and field consistency verification based on the village-level transaction knowledge graph, the to-do list set and the responsibility person mapping set, and generating a standardized report file set; receiving a document writing requirement, selecting a template structure and an element set based on the village-level transaction knowledge graph, the structured conference record and the standardized report file set, and generating a standardized document draft; receiving a village condition query request and a file upload task, outputting a query result based on the village-level transaction knowledge graph and the semantic indexing mechanism of the HSTF model and marking the information source path, and calling the version identification table and the node time index table to generate a file version record; adjusting the model parameters based on the task execution feedback information and updating the village-level transaction knowledge graph to generate a village-level transaction intelligent auxiliary result set.

2. The village-level transaction intelligent assistance method based on artificial intelligence according to claim 1, characterized in that, The generation of the report template set, the unified semantic representation structure and the semantic confidence matrix comprises: collecting conference voice, document text, report templates, policy files and basic data, performing format standardization and source marking, and converging to form a multi-source input information set; constructing and loading the HSTF model, the HSTF model comprising a semantic hierarchical coding unit, a context alignment unit, a voice recognition unit, a semantic reasoning unit, a semantic indexing mechanism and a self-supervised backflow unit; loading the parameter configuration of the semantic hierarchical coding unit and the context alignment unit, and establishing a processing channel for the multi-source input information set; calling the semantic hierarchical coding unit to perform hierarchical semantic coding on the multi-source input information set to generate hierarchical semantic representations, and calling the context alignment unit to perform cross-source context alignment and conflict resolution on the hierarchical semantic representations, and merging to obtain a unified semantic representation structure; based on the HSTF model, performing structure analysis and field extraction on the report templates in the multi-source input information set to extract a report template set; based on the calculation results of the HSTF model in the semantic hierarchical coding unit and the context alignment unit, calculating a semantic confidence matrix.

3. The method of claim 1, wherein the method is based on artificial intelligence. The generation of the village-level transaction knowledge graph, the version identification table and the node time index table comprises: determining entity extraction rules and relationship extraction rules according to the aligned semantic boundaries in the unified semantic representation structure and the confidence threshold in the semantic confidence matrix; extracting villager entities, meeting entities, policy entities, document entities, task entities and data field entities from the unified semantic representation structure according to the entity extraction rules to generate six types of entity nodes; generating membership relationships, initiation relationships, reference relationships, dependency relationships and source mapping relationships according to the relationship extraction rules to form a relationship mapping set; In the village-level transaction knowledge graph, population data nodes, land data nodes, and financial data nodes are constructed, the population data nodes are connected with villager entities through source mapping relationships, the land data nodes are connected with task entities through dependency relationships, and the financial data nodes are connected with document entities through reference relationships; A version identification table and a node time index table are established, the version states of the six types of entity nodes, the population data nodes, the land data nodes, and the financial data nodes are written into the version identification table, and the creation time and the update time of the nodes are written into the node time index table.

4. The village-level transaction intelligent assistance method based on artificial intelligence according to claim 1, characterized in that, The generation of the structured meeting record, the to-do list set, and the responsibility person mapping set includes: The collected meeting voice is input into a speech recognition unit of an HSTF model, acoustic pre-processing and dialect feature extraction are performed, and a speech feature stream is generated; Based on the speech feature stream, dialect speech transcription is performed, meeting transcription text is generated, semantic blocking is performed, and a semantic blocking fragment set is generated; Based on the village-level transaction knowledge graph, topic analysis, speech analysis, resolution analysis, and time constraint analysis are performed on the semantic blocking fragment set, and a structured meeting record is generated; Based on the structured meeting record, task descriptions, responsible subjects, and completion time limits are extracted, and a to-do list set and a responsibility person mapping set are generated.

5. The method of claim 1, wherein the method is based on artificial intelligence. The generation of the standardized report file set includes: The report template set is parsed and field constraint information is extracted, the field name, field type, value range, and filing rule are determined; The population data nodes, the land data nodes, and the financial data nodes in the village-level transaction knowledge graph are read, and a matching relationship between the fields and the data node attributes is established based on the field constraint information; Based on the to-do list set and the responsibility person mapping set, the filing object range is determined, and the matching relationship is limited to the node attribute set within the filing object range; According to the matching relationship, data filling is performed on the report template set, and a draft version of the standardized report file set is generated; Field consistency verification is performed on the draft version, the verification items include cross-field logic consistency, value range consistency, and source path consistency, and a verification result is obtained; The verification result is recorded as a report verification record and written into the village-level transaction knowledge graph, the draft version is finalized according to the verification result, and a standardized report file set is generated.

6. The village-level transaction intelligent assistance method based on artificial intelligence according to claim 1, characterized in that, The generation of the standardized report file set includes: The generation of the standardized report file set includes: The generation of the standardized report file set includes: The generation of the standardized report file set includes: The generation of the standardized report file set includes: The generation of the standardized report file set includes:

7. The method of claim 1, wherein the method is based on artificial intelligence. ​ Receiving village condition query request and file upload task, reading village-level transaction knowledge graph, version identification table and node time index table, calling semantic indexing mechanism of HSTF model; Based on the semantic indexing mechanism, the semantic analysis and condition mapping of the village condition query request are performed, the matching nodes and matching relationships in the village-level transaction knowledge graph are retrieved, the query results are generated, and the information source path is marked in the query results; Based on the version identification table and the node time index table, the file upload task is executed for deduplication, field comparison and version tracking, the matching results are classified into three states of new record, replacement record and retention record, and the corresponding version identification and time index information are recorded; According to the matching result, a file version record is generated, and the version identification, creation time, update time and information source path are recorded in the file version record.

8. The village-level transaction intelligent assistance method based on artificial intelligence according to claim 1, characterized in that, The generation of the village-level transaction intelligent auxiliary result set includes: Receiving and gathering task execution feedback information, the task execution feedback information includes confirmation results of structured meeting records, confirmation results of standardized report files, confirmation results of report verification records, confirmation results of standardized document preliminary drafts, confirmation results of document reference indexes, external execution states of query results and external execution states of file version records; Input the task execution feedback information into the self-supervised backflow unit of the HSTF model, sort and align according to the source label and the time label, and complete the input preparation for parameter adjustment; Call the self-supervised backflow unit to adjust the HSTF model parameters according to the sorted and aligned task execution feedback information, and obtain the adjusted HSTF model parameters; Based on the task execution feedback information, update the entity confidence and relationship weight in the village-level transaction knowledge graph, write the update result back to the village-level transaction knowledge graph, and keep consistent with the version identification table and the node time index table; Based on the updated village-level transaction knowledge graph, standardized report files, standardized document preliminary drafts, query results and file version records, call the self-supervised backflow unit of the HSTF model to perform result aggregation operation, and generate a village-level transaction intelligent auxiliary result set containing task completion state, data consistency evaluation, information traceability path and model parameter update summary.

9. An artificial intelligence-based intelligent auxiliary system for village-level affairs, which executes an artificial intelligence-based intelligent auxiliary method for village-level affairs according to any one of claims 1 to 8. It includes: A multi-source data acquisition module for acquiring multi-source data, performing format standardization and generating a multi-source input information set; An HSTF model module for building and loading an HSTF model, the model including a semantic hierarchical coding unit, a context alignment unit, a speech recognition unit, a semantic reasoning unit, a semantic indexing mechanism and a self-supervised backflow unit; A village-level transaction knowledge graph construction module for extracting entity information based on a unified semantic representation structure and a semantic confidence matrix, establishing multi-dimensional relationships and constructing a village-level transaction knowledge graph; A meeting record generation module for performing dialect speech transcription and semantic segmentation through a speech recognition unit to generate structured meeting records, to-do list sets and responsibility person mapping sets; A report generation and verification module for performing field matching, data filling and consistency verification based on the village-level transaction knowledge graph and the responsibility person mapping set to generate a standardized report file set; The document generation module is configured to invoke the semantic reasoning unit, combine the knowledge graph, the conference record and the report file to generate a standardized document preliminary draft and a document reference index; The query and file management module is configured to output a query result based on a semantic index mechanism, perform file uploading, deduplication and version tracking, and generate a file version record; The intelligent feedback and optimization module is configured to gather task execution feedback information, invoke a self-supervised backflow unit to adjust model parameters and update the knowledge graph, and generate a village-level intelligent auxiliary result set.

Images