Government affair service combing calculation method and device and readable storage medium

Through the calculation method of government service sorting and calculation, knowledge graphs and large-model technology are used to automatically process government service call records, solving the problem of low manual sorting efficiency, realizing the optimization of government service processes and scientific decision-making.

CN119066207APending Publication Date: 2024-12-03CAPE DIGITAL TECH (GUANGDONG) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411168007.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing government service processing information needs to be sorted out by manual standards, which is inefficient and prone to errors, resulting in insufficient administrative efficiency.

Method used

The government service sorting calculation method is used to build a knowledge graph by collecting call records, and the graph neural network and large-scale model technology are used to automatically identify and extract key information to form standardized processing data, and display it to the end user through the knowledge graph graph graphic tool.

Benefits of technology

Optimize government service processes, improve work efficiency, improve policy and service accuracy, reduce the burden of manual judgment, and realize data-driven decision-making.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066207A_ABST
    Figure CN119066207A_ABST
Patent Text Reader

Abstract

The invention discloses a government affair service carding method and device and a storage medium, and mainly constructs and fuses a knowledge graph of transaction items by collecting and analyzing call records and document handling process data of government affair services. According to the method, a knowledge graph is processed by adopting a graph neural network technology, and classification is carried out by taking an identity card of a clerk / a code of an enterprise organization as a center. According to the technical scheme, natural person or enterprise classification is performed on knowledge graph data through a large-model text understanding technology, and the knowledge graph data is displayed by using a graphical tool, so that the intelligence and precision of government affair services are improved, the government service efficiency is improved, and the citizen and enterprise experience is optimized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a data combing method, and in particular to a government service combing calculation method, device and readable storage medium. Background Art

[0002] With the accelerated advancement of standardization, regularization and facilitation of government services, the collection of information relies on two points. First, the unification of the elements of the implementation list of government service items lays the foundation for the unification of item coding and names. Second, to promote the standardization of government services, the government service approval department should simultaneously push the relevant government service item processing information and results to the industry regulatory department.

[0003] The processing information model constructed above requires manual standardization and uniform sorting of processing information due to the different designs of various business systems, forming records of "service code", "service item name", "service person ID card / enterprise organization code", and "processing time". Therefore, at this stage, efficient government services continue to require effective sorting of items to improve administrative efficiency, so as to achieve the result of filling out a form, submitting a set of materials, and submitting them all at once. At present, the items that need to be sorted out are positively sorted out by the leading unit and the relevant competent departments, and the business-savvy personnel rely on their business knowledge in their departments. For example, in order to efficiently handle the birth of a child, the hospital, human resources and social security departments, and public security departments will be involved after the birth of the child. This is manually sorted out based on common sense + business knowledge. Summary of the invention

[0004] In order to solve the deficiencies of the above-mentioned technologies, the present invention provides a government service combing and calculation method, device and readable storage medium.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: a government service sorting and calculation method, comprising the following steps:

[0006] Collect call records of government services and use them to build a knowledge graph of the content of the affairs to be handled;

[0007] Use the data of the handling process to build a knowledge graph of the handling matters;

[0008] The graph-based neural network integrates the knowledge graph of the above-constructed items and the knowledge graph of the items to be handled to form a knowledge graph data with "personnel ID card / enterprise organization code" as the center, the second-level node is "items to be handled", and the third-level node is "handling time";

[0009] Using the text understanding of the big model, the knowledge graph data is classified into "natural persons" or "enterprises";

[0010] Use knowledge graph graphical tools to display knowledge graph data graphically for end users to use.

[0011] Furthermore, call records of government services are collected, and the call records are used to construct a knowledge graph of the content of the matters to be handled, including: collecting call records of government services, converting them through a large model with a voice conversion function to obtain recorded text; identifying key information in the recorded text; using knowledge graph technology to set the handling time, using the large model to read the data after key information identification, collecting relationships, and forming a knowledge graph with three-level nodes; using the RAG technology of the large model to match the matters to be handled according to the service guide of the government services, manually intervening to compare the results, and obtaining knowledge graph data results that meet the standards.

[0012] Furthermore, call records of government services are collected, including: call records of window staff and front desk staff in physical service halls, and call records of 12345 operators.

[0013] Furthermore, key information is identified in the recorded text, including: using the text understanding of the big model to import all call records, perform entity recognition, event extraction, and keyword extraction, and identify the "mobile phone number of the person handling the matter", "content of the matter being handled", "name of the matter being handled", and "handling time" respectively.

[0014] Furthermore, key information of the recorded text is identified; the processing time is set using knowledge graph technology, and the data after key information identification is read using a large model to collect relationships to form a knowledge graph with three-level nodes, including: using knowledge graph technology to set "processing time" as a relationship point within a week, using a large model to read data, and collecting relationships to form a knowledge graph with "mobile phone number of the person handling the matter" as the center, the second-level node being "content of the matter being handled", and the third-level node being "processing time", and the second-level node "content of the matter being handled" is set to be greater than 3.

[0015] Furthermore, the RAG technology of the big model is used to match service items according to the service guide of the government service, and manual intervention is performed to compare the results to obtain knowledge graph data results that meet the standards, including: using the big model RAG technology to perform enhanced retrieval on the "content of service items" in the knowledge graph to accurately match the standard "name of service items", and manual intervention is performed during this period to compare the results to obtain knowledge graph data results that meet the standards.

[0016] Furthermore, the data of the case handling process includes the use of large-scale model text understanding to identify entities such as “service code”, “service item name”, “service person ID card / enterprise organization code” and “processing time”.

[0017] Furthermore, the secondary node "matters to be handled" is set to be greater than 3.

[0018] A government service combing device, the device comprising: a memory, a processor, and a government service combing program stored in the memory and executable on the processor, the government-related data combing program being configured to implement the steps of the government service combing algorithm method as claimed in any one of claims 1 to 8.

[0019] A computer-readable storage medium stores a government service combing program, which, when executed by a processor, implements the steps of the government service combing calculation method as claimed in any one of claims 1 to 8.

[0020] The present invention discloses a government service combing calculation method, device and readable storage medium, which have the following advantages:

[0021] Optimize government service processes: By collecting and analyzing call record data for government services and combining it with big model technology, key information can be automatically identified and extracted, thereby optimizing the government service process.

[0022] Improve work efficiency: Traditional government affairs sorting work usually relies on manual work, which is inefficient and prone to errors. This patented method greatly reduces the workload of manual judgment and improves work efficiency by automatically processing a large amount of unstructured data.

[0023] Improving the accuracy of policies and services: Using big data analysis results, government departments can give more practical recommendations on "one thing" related policies and services based on local conditions, such as providing one-stop services in cities with high population mobility.

[0024] Enhanced data understanding and reasoning capabilities: Through the construction of knowledge graphs and the application of technologies such as graph neural networks, large models can better understand the entities and background information in the text, thereby enhancing data processing and reasoning capabilities.

[0025] Achieve data-driven decision-making: This patent uses a data-driven approach to optimize business processes, making government service decisions more scientific and well-founded.

[0026] Graphical display of knowledge graphs: Through graphical tools such as Arc2, knowledge graph data can be presented to end users in an intuitive way, making complex data easier to understand and use. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flowchart of the algorithm method for sorting out government services for this application. DETAILED DESCRIPTION

[0028] The present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0029] like Figure 1 The government service combing calculation method shown includes the following steps:

[0030] Collect call records of government services and use them to build a knowledge graph of the content of the affairs to be handled;

[0031] Collect call records of government services, including: call records of window staff and front desk staff in physical service halls, call records of 12345 operators, and other unstructured data.

[0032] Collect call records of government services and convert them through a large model with speech conversion function to obtain recorded text;

[0033] For example, using Cape Cloud's Kaiwu big model is one of the basic capabilities of the current big model. The speech conversion process is consistent with the existing technology, such as iFlytek's speech conversion.

[0034] Identify key information in recorded text;

[0035] Exemplary, key information is identified in the recorded text, including: using the text understanding of the large model to import all call records, and performing entity recognition, event extraction, and keyword extraction to identify the "mobile phone number of the person handling the matter", "content of the matter being handled", "name of the matter being handled", and "handling time" respectively.

[0036] Using the Kaiwu big model of Cape Cloud, the big model's "information extraction" or "key information identification" is used, which is a subfield of natural language processing (NLP). Information extraction can automatically identify and extract specific types of entities (such as names, places, dates, etc.), relationships between entities, and event information from text. The process of extracting information is as follows. For example, for a conversation about consulting how to apply for a real estate certificate, there are the following steps:

[0037] The large model uses Named Entity Recognition (NER): it identifies named entities such as names of people, places, and organizations in the text, and identifies the mobile phone numbers, titles, and even names of people who may be mentioned in the conversation.

[0038] The large model uses relation extraction to identify the relationship between entities, such as "working relationship", "geographical location relationship", etc., and to identify that in a conversation, for example, "I want to apply for a real estate certificate", it is a service relationship rather than a complaint relationship.

[0039] The large model uses event argument extraction (Argument Extraction): identifying entities related to the event, such as the participants, time, and location of the event, mainly extracting the time, location, etc. in the record.

[0040] The large model uses summary generation (Summarization): generating short and concise summaries from long text conversations, retaining the main points and key information, thereby forming the "content of the matter being handled".

[0041] Use knowledge graph technology to set processing time, use big models to read data after key information identification, collect relationships, and form a knowledge graph with three-level nodes; use big model's RAG technology to match service items according to the government service guide, manually intervene to compare results, and obtain knowledge graph data results that meet standards.

[0042] Use knowledge graph technology to set processing time, use big models to read data after key information identification, collect relationships, and form a knowledge graph with three-level nodes, including: use knowledge graph technology to set "processing time" as a relationship point within a week, use big models to read data, collect relationships, and form a knowledge graph with "mobile phone number of the person doing the work" as the center, the second-level node is "content of the matter being handled", and the third-level node is "processing time", and the second-level node "content of the matter being handled" is set to be greater than 3.

[0043] Among them, Knowledge Graphs (KGs) is a data structure used to store and represent knowledge. It is usually designed in the form of nodes and edges, where nodes represent entities (such as people, places, concepts, etc.) and edges represent the relationship between these entities. Knowledge graphs play an important role in large language models and can be used to enhance the model's understanding and reasoning capabilities. The key features of knowledge graphs are:

[0044] Entities and relationships, each entity has a unique identifier, and entities are connected by clear relationships. Here, "the person who handles the matter mobile phone number" is the entity, and the second-level and third-level nodes are its first-level and second-level relationships.

[0045] Structured data: Knowledge graphs organize information in a structured form, making it easier for machines to understand and process the information. Mind maps are a form of structured data.

[0046] Key technologies of knowledge graph:

[0047] Embedding method: By mapping entities and relations into a low-dimensional vector space, similar entities are placed closer in the vector space, which helps the model understand and reason about the relationships between entities.

[0048] Graph Neural Networks (GNNs): This is a deep learning model specifically designed to process graph data. It can capture complex relationships between entities and propagate information by iteratively updating node representations.

[0049] Path ranking algorithm: This method predicts the relationship between entities by learning a scoring function on different paths.

[0050] Graph Convolutional Networks (GCNs): A special GNN architecture that can effectively aggregate the information of neighboring nodes to update the representation of the central node.

[0051] Graph Attention Mechanism: Similar to the attention mechanism, but specifically for graph data, it can highlight important parts of the graph.

[0052] In large models, the knowledge graph in this patent mainly enhances understanding capabilities: by accessing entities and relationships in the knowledge graph, the model can better understand the entities and background information in the text.

[0053] The RAG technology of the big model is used to match service items according to the service guide of the government service, and manual intervention is used to compare the results to obtain knowledge graph data results that meet the standards, including: using the RAG technology of the big model to perform enhanced retrieval on the "content of service items" in the knowledge graph to accurately match the standard "name of service items", and manual intervention is used during this period to compare the results to obtain knowledge graph data results that meet the standards.

[0054] For example, the big model must first understand the service guide. The enlightened big model learns on the training corpus, which is the thousands of service guides on the current government service network. All the service guides on the government service network are collected into a document and given to the enlightened big model for understanding. Then, the big model's RAG technology is used to input the "content of the items to be handled", which is to ask the big model a question. The question is pieced together like this: "If there is a sentence that describes it like this: #content of the items to be handled#, which item is it?" Then the big model will search and answer. The big model has an interface for the above process. We can use the interface in a programming way to traverse the "content of the items to be handled" in the knowledge graph to obtain the corresponding results.

[0055] Among them, RAG (Retrieval-Augmented Generation) is a method that combines retrieval technology and language models (usually large models) to consider information in external knowledge bases when generating responses.

[0056] Use the data of the handling process to build a knowledge graph of the handling matters;

[0057] Processing data, including the use of large-scale model text understanding to identify entities such as "service code", "service item name", "service provider ID card / enterprise organization code" and "processing time"

[0058] Using knowledge graph technology, set "processing time" as a relationship point within a week, use the big model to read data, collect relationships, and form a knowledge graph with "personnel ID card / company organization code / mobile phone number" as the center, the second-level node as "processing items", and the third-level node as "processing time", and set the sub-node "processing items" to be greater than 3.

[0059] The graph-based neural network integrates the knowledge graph of the above-constructed items and the knowledge graph of the items to be handled to form a knowledge graph data with "personnel ID card / enterprise organization code" as the center, the second-level node is "items to be handled", and the third-level node is "handling time";

[0060] Among them, graph neural networks (GNNs) are a technology in knowledge graphs. Two knowledge graphs are fused based on the center point and the corresponding primary and secondary relationships. For example, the center point is a mobile phone number in one graph and a mobile phone number in the other. The simplest fusion is to use a complete match for fusion, but based on the understanding of the matter by the big model, it will determine whether the primary relationship is an enterprise-related matter or a natural person matter, and then two types of central nodes will be formed: "personnel ID card-mobile phone number" and "enterprise organization code-mobile phone number". When fusion is carried out, semantically similar or unified formats will be used for conversion. For example, "handling time" may have different formats, which will be automatically converted, and the path ranking algorithm will be automatically used to retain the top 3.

[0061] Using the text understanding of the big model, the knowledge graph data is classified into "natural persons" or "enterprises";

[0062] The knowledge graph data is graphically displayed using a knowledge graph graphical tool for end users to use. The tool used for knowledge graph graphical display is Arc2.

[0063] In summary, this patent combines big data processing technology to conduct in-depth analysis of data such as processing results and consultation results. This method can help relevant departments break out of traditional thinking patterns and formulate relevant policies and services for "one thing" more flexibly according to local actual conditions. For example, in cities with high population mobility, residents may soon need to go through relevant procedures such as opening a restaurant or a small shop after completing the household registration procedures. In this case, relevant departments can work together to provide citizens with one-stop services to handle all relevant matters at one time. The main advantages of this patent include: first, a data-driven approach is used to optimize business processes; second, the workload of manual judgment is greatly reduced; third, the results of business combing can be made more accurate and convenient.

[0064] In addition, the present application also discloses a government service combing device, which includes: a memory, a processor, and a government service combing program stored in the memory and executable on the processor, and the government-related data combing program is configured to implement steps such as a government service combing algorithm method.

[0065] A computer-readable storage medium is also proposed, on which a government service combing program is stored. When the government service combing program is executed by a processor, the steps of the government service combing calculation method are implemented.

[0066] The above implementation modes are not limitations of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the scope of the technical solution of the present invention also belong to the protection scope of the present invention.

Claims

1. A government service sorting and calculation method, characterized in that: The following steps are involved: Collect call records of government services and use them to build a knowledge graph of the content of the affairs to be handled; Use the data of the handling process to build a knowledge graph of the handling matters; The graph-based neural network integrates the knowledge graph of the above-constructed items and the knowledge graph of the items to be handled to form a knowledge graph data with "personnel ID card / enterprise organization code" as the center, the second-level node is "items to be handled", and the third-level node is "handling time"; Using the text understanding of the big model, the knowledge graph data is classified into "natural persons" or "enterprises"; Use knowledge graph graphical tools to display knowledge graph data graphically for end users to use.

2. The government service sorting and calculation method according to claim 1 is characterized in that: The method of collecting call records of government services and using call records to construct a knowledge graph of the content of matters to be handled includes: collecting call records of government services, converting them through a large model with a voice conversion function to obtain record text; identifying key information in the record text; setting the handling time using knowledge graph technology, using a large model to read the data after key information identification, collecting relationships, and forming a knowledge graph with three-level nodes; using the RAG technology of the large model to match matters to be handled according to the service guide of the government services, manually intervening to compare the results, and obtaining knowledge graph data results that meet the standards.

3. The government service sorting and calculation method according to claim 2 is characterized in that: The call records collected for government services include: call records of window staff and front desk staff at physical service halls, and call records of 12345 operators.

4. The government service sorting and calculation method according to claim 2 is characterized in that: The key information identification of the recorded text includes: using the text understanding of the large model to import all call records, and performing entity recognition, event extraction, and keyword extraction to identify the "mobile phone number of the person handling the matter", "content of the matter being handled", "name of the matter being handled", and "handling time" respectively.

5. The government service sorting and calculation method according to claim 4 is characterized in that: The key information of the record text is identified; the processing time is set using knowledge graph technology, and the data after key information identification is read using a large model to collect relationships to form a knowledge graph with three levels of nodes, including: using knowledge graph technology to set "processing time" as a relationship point within a week, using a large model to read data, and collecting relationships to form a knowledge graph with "mobile phone number of the person handling the matter" as the center, the second-level node is "content of the matter being handled", and the third-level node is "processing time", and the second-level node "content of the matter being handled" is set to be greater than 3.

6. The government service combing and calculation method according to claim 5 is characterized in that: The RAG technology using a large model matches service items according to the service guide of the government service, and manually intervenes to compare the results to obtain knowledge graph data results that meet the standards, including: using the RAG technology using a large model to perform enhanced retrieval on the "content of service items" in the knowledge graph to accurately match the standard "name of service items", and manually intervenes to compare the results during this period to obtain knowledge graph data results that meet the standards.

7. The government service sorting and calculation method according to claim 1 is characterized in that: The handling process data includes using the big model's text understanding to identify the entities "service code", "service item name", "service person's ID card / enterprise organization code" and "handling time".

8. The government service sorting and calculation method according to claim 1 is characterized in that: Set the secondary node "matters to be handled" to be greater than 3.

9. A government service sorting device, characterized in that: The device includes: a memory, a processor, and a government service combing program stored in the memory and executable on the processor, wherein the government-related data combing program is configured to implement the steps of the government service combing algorithm method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a government service combing program, which, when executed by a processor, implements the steps of the government service combing calculation method as described in any one of claims 1 to 8.