Method and system for realizing accurate pushing of tax information
Through in-depth analysis of taxpayer data and intelligent processing of DeepSeek model, the problem of lack of targeted traditional tax publicity methods is solved, the precise push of tax information is achieved, the accuracy and efficiency of information transmission is improved, and tax risks and tax costs are reduced.
Patent Information
- Application Number
- CN202510416091.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-03
- Publication Date
- 2025-07-11
AI Technical Summary
The lack of targeted traditional tax publicity methods makes it difficult for taxpayers to quickly find information related to their own business, increasing tax costs and tax risks.
Through the collection and preprocessing of taxpayer data, the construction of tax information knowledge base, the optimization and adaptation of the DeepSeek model, and the construction of taxpayer portraits, the large language model DeepSeek is used to accurately match and push tax information.
It realizes the accurate push of tax information, improves the accuracy and efficiency of information communication, reduces tax risks and tax costs for taxpayers, and improves taxpayers' satisfaction.
Smart Images
Figure CN120296218A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the fields of artificial intelligence and tax management, and specifically provides a method and system for realizing accurate push of tax information. Background Art
[0002] With the rapid economic development and the increasing complexity of taxes, enterprises and taxpayers are facing the challenge of obtaining applicable taxes in a timely and accurate manner. Traditional tax publicity methods are often extensive and lack pertinence, resulting in taxpayers having difficulty quickly finding information related to their own businesses, increasing the tax payment costs and tax risks.
[0003] The development of artificial intelligence technology provides a new way to solve this problem. Large language models can process and understand massive amounts of text information. By analyzing and learning tax policy texts, it is expected to achieve accurate policy push for taxpayers. As a new generation of powerful language models, DeepSeek models have excellent reasoning capabilities and the ability to understand complex semantics, and have shown great application potential in multiple fields. Applying them to the field of tax policy push can effectively improve the accuracy and efficiency of push.
[0004] How to accurately match and push tax information that meets the needs of taxpayers through the analysis of taxpayer data and the intelligent processing of DeepSeek models, improve the efficiency and accuracy of tax information transmission, and reduce the tax risks and tax payment costs of taxpayers is an urgent problem to be solved in this field. Summary of the Invention
[0005] The present invention aims at the deficiencies of the above-mentioned existing technologies and provides a method for realizing accurate push of tax information with strong practicability.
[0006] A further technical task of the present invention is to provide a system for realizing accurate push of tax information with reasonable design, safety and applicability.
[0007] The technical solution adopted by the present invention to solve its technical problems is as follows:
[0008] A method for realizing accurate push of tax information has the following steps:
[0009] S1. Collection and preprocessing of taxpayer data;
[0010] S2. Construction of a tax information knowledge base;
[0011] S3. Optimization and adaptation of the DeepSeek model;
[0012] S4. Construction of taxpayer portraits;
[0013] S5. Matching and push of information.
[0014] Further, in step S1, collect the basic information of taxpayers, where the basic information includes enterprise type, business scope, scale, and registration time;
[0015] Collect the tax declaration data of taxpayers, clean and preprocess the collected data, remove noise data and outliers, and convert the data into a format suitable for model processing.
[0016] Further, in step S2, collect and organize various tax documents, use text parsing technology to structure the information documents, extract key information, store the structured information in the knowledge base, and establish an index for quick retrieval.
[0017] Further, in step S3, fine-tune the DeepSeek model according to the characteristics of taxation. Train the model with a large amount of tax texts and related cases to make it understand tax professional terms and semantics;
[0018] Optimize the parameter settings of the model to improve the model's performance in tax policy understanding and matching tasks;
[0019] Establish an interface between the model and the tax knowledge base to enable the model to query and call the information in the knowledge base.
[0020] Further, in step S4, construct a taxpayer profile based on the basic information and tax declaration data of taxpayers using data mining and machine learning algorithms;
[0021] The taxpayer profile includes the industry characteristics, tax payment behavior characteristics, and risk preferences of taxpayers to depict the characteristics of taxpayers.
[0022] Further, in step S5, input the taxpayer profile into the optimized DeepSeek model, and the model matches and filters in the tax information knowledge base according to the characteristics of taxpayers;
[0023] Utilize the reasoning ability of the model to analyze the tax information applicable to taxpayers and sort it according to relevance and importance;
[0024] Precisely push the matched tax policies to taxpayers through multiple channels.
[0025] A system for realizing precise push of tax information. First, for taxpayer data collection and preprocessing, collect the basic information of taxpayers, where the basic information includes enterprise type, business scope, scale, and registration time;
[0026] Collect the tax declaration data of taxpayers, clean and preprocess the collected data, remove noise data and outliers, and convert the data into a format suitable for model processing.
[0027] Then, construct the tax information knowledge base, optimize and adapt the DeepSeek model;
[0028] Finally, construct the taxpayer portrait. Based on the taxpayer's basic information and tax declaration data, use data mining and machine learning algorithms to construct the taxpayer portrait, and perform information matching and push.
[0029] Furthermore, in the construction of the tax information knowledge base, collect and organize various tax documents, use text parsing technology to structure the information documents, extract key information, store the structured information in the knowledge base, and establish an index for quick retrieval.
[0030] Furthermore, in the optimization and adaptation of the DeepSeek model, fine-tune the DeepSeek model according to the characteristics of taxation. Use a large number of tax texts and relevant cases to train the model so that it understands tax professional terms and semantics;
[0031] Optimize the parameter settings of the model to improve the model's performance in tax policy understanding and matching tasks;
[0032] Establish an interface between the model and the tax knowledge base so that the model can query and call the information in the knowledge base.
[0033] Furthermore, in the information matching and push, input the taxpayer portrait into the optimized DeepSeek model. The model matches and filters in the tax information knowledge base according to the taxpayer's characteristics;
[0034] Use the reasoning ability of the model to analyze the tax information applicable to taxpayers and sort it according to relevance and importance;
[0035] Precisely push the matched tax policies to taxpayers through multiple channels.
[0036] A method and system for realizing precise push of tax information according to the present invention have the following outstanding beneficial effects compared with the prior art:
[0037] (1) Through in-depth analysis of taxpayer data and intelligent processing of the DeepSeek model, the present invention can accurately match the tax information applicable to taxpayers, avoid the blindness of traditional push methods, and improve the accuracy of information transmission.
[0038] (2) The model can quickly process a large amount of taxpayer data and tax information, realize real-time push of information, greatly shorten the time for taxpayers to obtain tax information, and improve the efficiency of tax services.
[0039] (3) Taxpayers can obtain accurate tax information in a timely manner, which helps them correctly fulfill their tax obligations and reduce tax risks caused by inaccurate understanding.
[0040] (4) Precise push can improve taxpayers' satisfaction with the services of the tax department and promote the harmonious development of tax relations. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.
[0042] Appendix Figure 1 is a schematic flowchart of a method for realizing precise push of tax information. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0043] In order to enable those skilled in the art of this technology to better understand the solution of the present invention, the following will further elaborate on the present invention in conjunction with specific embodiments. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.
[0044] The following gives a preferred embodiment:
[0045] As Figure 1 shown, a method for realizing precise push of tax information in this embodiment has the following steps:
[0046] S1. Collection and preprocessing of taxpayer data;
[0047] Collect the basic information of taxpayers, including enterprise type, business scope, scale, registration time, etc.
[0048] Collect the tax declaration data of taxpayers, such as tax types, tax amounts, declaration periods, etc.
[0049] Clean and preprocess the collected data, remove noise data and outliers, and convert the data into a format suitable for model processing.
[0050] Through the tax system interface, regularly obtain the basic information and tax declaration data of taxpayers, and use data cleaning algorithms to clean the collected data, such as removing duplicate data, correcting incorrect data formats, etc. Classify and code the business scope of taxpayers for subsequent data analysis and model processing.
[0051] S2. Construction of a tax information knowledge base;
[0052] Collect and organize various tax documents, including laws and regulations, departmental rules, local policies, etc.
[0053] Use text parsing technology to structure the documents and extract key information such as applicable objects, content, preferential conditions, implementation time, etc.
[0054] Store the structured information in a knowledge base and establish an index for quick retrieval.
[0055] Arrange for professional staff to collect the latest tax documents, including issued announcements, implemented rules, etc.
[0056] Use text extraction tools in natural language processing technology to extract key information from the documents, such as names, document numbers, release dates, applicable scopes, content, etc.
[0057] Store the extracted information in a relational database, establish the association between the ID and various information, and at the same time establish a full-text index for the content for convenient and quick retrieval.
[0058] S3. Optimization and adaptation of the DeepSeek model;
[0059] According to the characteristics of the tax field, fine-tune the DeepSeek model. Use a large number of tax texts and related cases to train the model so that it can better understand tax professional terms and semantics.
[0060] Optimize the parameter settings of the model to improve the performance of the model in tax understanding and matching tasks.
[0061] Establish an interface between the model and the tax knowledge base so that the model can quickly query and call the information in the knowledge base.
[0062] Sort out a large number of tax texts and related cases from public data sets and internal data of tax departments as training data for model fine-tuning.
[0063] Use a deep learning framework, load the pre-trained DeepSeek model, and input the training data into the model for fine-tuning. During the fine-tuning process, adjust the hyperparameters of the model, such as the learning rate, number of iterations, etc., to improve the performance of the model.
[0064] Develop an interface program between the model and the tax knowledge base so that the model can query the information in the knowledge base through the interface and conduct reasoning and analysis based on the query results.
[0065] S4. Construction of taxpayer portraits;
[0066] According to the basic information of taxpayers, tax declaration data, and other relevant data, use data mining and machine learning algorithms to construct taxpayer portraits.
[0067] The taxpayer profile includes dimensions such as the taxpayer's industry characteristics, tax payment behavior characteristics, risk preferences, etc., comprehensively depicting the characteristics of taxpayers.
[0068] Using clustering algorithms, conduct clustering analysis on the basic information and tax declaration data of taxpayers, and classify taxpayers into different groups, such as large enterprises, medium-sized enterprises, small enterprises, individual industrial and commercial households, etc.
[0069] For each group, analyze its tax payment behavior characteristics, such as the change trend of tax payment amounts, the stability of the declaration cycle, etc., and establish corresponding characteristic indicators.
[0070] According to the industry characteristics of taxpayers, analyze the possible tax risks they may face, such as the risks of applying tax incentives and invoice management risks that may exist in certain industries, and incorporate the risk indicators into the taxpayer profile.
[0071] S5. Matching and pushing information;
[0072] Input the taxpayer profile into the optimized DeepSeek model, and the model conducts matching and screening in the tax knowledge base according to the characteristics of taxpayers.
[0073] Utilize the reasoning ability of the model to analyze the taxes that taxpayers may be applicable to, and sort them according to relevance and importance.
[0074] Push the matched taxes to taxpayers through various channels, such as WeChat, SMS, email, etc.
[0075] When new tax information is released, input the tax information into the optimized DeepSeek model, and the model automatically screens out the list of taxpayers who may be applicable to this taxpayer according to the content of the tax information and the taxpayer profile.
[0076] For the screened taxpayers, generate personalized information push content according to the matching degree and importance calculated by the model, including the original text, interpretation, description of applicable conditions, etc.
[0077] Push the information to taxpayers through the message push block of the tax system. When taxpayers log in to the tax system, they can see the pushed tax information at a prominent position and can click to view the detailed content. At the same time, for important information, taxpayers can also be reminded via SMS or email.
[0078] Based on the above method, in one implementation of the system for accurately pushing tax information in this embodiment, first, collect and preprocess taxpayer data, and collect the basic information of taxpayers. The basic information includes enterprise type, business scope, scale, and registration time;
[0079] Collect the tax declaration data of taxpayers, clean and preprocess the collected data, remove noise data and outliers, and convert the data into a format suitable for model processing.
[0080] Then, construct the tax information knowledge base, optimize and adapt the DeepSeek model;
[0081] Finally, construct the taxpayer portrait. Based on the basic information and tax declaration data of taxpayers, use data mining and machine learning algorithms to construct the taxpayer portrait, and perform information matching and pushing.
[0082] Among them, in the construction of the tax information knowledge base, collect and sort out various tax documents, use text parsing technology to structure the information documents, extract key information, store the structured information in the knowledge base, and establish an index for quick retrieval.
[0083] In the optimization and adaptation of the DeepSeek model, fine-tune the DeepSeek model according to the characteristics of taxation. Use a large number of tax texts and related cases to train the model so that it understands tax professional terms and semantics;
[0084] Optimize the parameter settings of the model to improve the performance of the model in tax policy understanding and matching tasks;
[0085] Establish an interface between the model and the tax knowledge base to enable the model to query and call the information in the knowledge base.
[0086] In the information matching and pushing, input the taxpayer portrait into the optimized DeepSeek model. The model matches and filters in the tax information knowledge base according to the characteristics of taxpayers;
[0087] Use the reasoning ability of the model to analyze the tax information applicable to taxpayers and sort it according to relevance and importance;
[0088] Precisely push the matched tax policies to taxpayers through various channels.
[0089] The above specific implementation manners are only specific cases of the present invention. The patent protection scope of the present invention includes but is not limited to the above specific implementation manners. Any technical solution that conforms to the technical solution recorded in the claims of the present invention and any appropriate changes or substitutions made by those of ordinary skill in the art shall fall within the patent protection scope of the present invention.
[0090] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A method for realizing accurate push of tax information, characterized in that, It has the following steps: S1. Taxpayer data collection and preprocessing; S2. Construction of the tax information knowledge base; S3. DeepSeek model optimization and adaptation; S4. Taxpayer profile construction; S5. Information matching and push.
2. The method for accurately pushing tax information according to claim 1, wherein In step S1, collect the basic information of taxpayers, where the basic information includes enterprise type, business scope, scale, and registration time; Collect the tax declaration data of taxpayers, clean and preprocess the collected data, remove noise data and outliers, and convert the data into a format suitable for model processing.
3. The method for accurately pushing tax information according to claim 2, characterized in that In step S2, collect and organize various tax documents, use text parsing technology to structure the information documents, extract key information, store the structured information in the knowledge base, and establish an index for quick retrieval.
4. A method for realizing accurate push of tax information according to claim 3, characterized in that, In step S3, fine-tune the DeepSeek model according to the characteristics of taxation. Use a large amount of tax texts and relevant cases to train the model so that it understands tax professional terms and semantics; Optimize the parameter settings of the model to improve the performance of the model in tax policy understanding and matching tasks; Establish an interface between the model and the tax information knowledge base to enable the model to query and call the information in the knowledge base.
5. The method for precisely pushing tax information according to claim 4, wherein In step S4, construct a taxpayer profile based on the basic information and tax declaration data of taxpayers, using data mining and machine learning algorithms; The taxpayer profile includes the industry characteristics, tax payment behavior characteristics, and risk preferences of taxpayers to depict the characteristics of taxpayers.
6. The method for realizing accurate push of tax information according to claim 5, wherein, In step S5, input the taxpayer profile into the optimized DeepSeek model. The model matches and filters in the tax information knowledge base according to the characteristics of taxpayers; Utilize the reasoning ability of the model to analyze the tax information applicable to taxpayers and sort it according to relevance and importance; Precisely push the matched tax policies to taxpayers through various channels.
7. A system for realizing precise push of tax information, characterized in that, First, for taxpayer data collection and preprocessing, collect the basic information of taxpayers, where the basic information includes enterprise type, business scope, scale, and registration time; Collect the tax declaration data of taxpayers, clean and preprocess the collected data, remove noise data and outliers, and convert the data into a format suitable for model processing. Then, conduct the construction of the tax information knowledge base and DeepSeek model optimization and adaptation; Finally, for taxpayer profile construction, construct a taxpayer profile based on the basic information and tax declaration data of taxpayers, using data mining and machine learning algorithms, and information matching and push.
8. A system for accurately pushing tax information according to claim 7, characterized in that, In the construction of the tax information knowledge base, collect and organize various tax documents, use text parsing technology to structure the information documents, extract key information, store the structured information in the knowledge base, and establish an index for quick retrieval.
9. The system for precisely pushing tax information according to claim 8, wherein In DeepSeek model optimization and adaptation, fine-tune the DeepSeek model according to the characteristics of taxation. Use a large amount of tax texts and relevant cases to train the model so that it understands tax professional terms and semantics; Optimize the parameter settings of the model to improve the performance of the model in tax policy understanding and matching tasks; Establish an interface between the model and the tax knowledge base to enable the model to query and call information in the knowledge base.
10. A system for precisely pushing tax information according to claim 9, characterized in that, In information matching and pushing, input the taxpayer portrait into the optimized DeepSeek model. The model matches and filters in the tax information knowledge base according to the characteristics of the taxpayer. Utilize the reasoning ability of the model to analyze the tax information applicable to taxpayers and sort it according to relevance and importance. Precisely push the matched tax policies to taxpayers through multiple channels.