Data analysis methods, devices, computer equipment, and media based on artificial intelligence

Abstract

This application belongs to the field of artificial intelligence technology and relates to a data parsing method, apparatus, computer equipment, and storage medium based on artificial intelligence. The method includes: acquiring policy data from a preset data source; cleaning the policy data to obtain corresponding first policy data; scoring the first policy data to obtain score data; filtering target policy data that meets preset high-score criteria from the first policy data based on the score data; acquiring preset target prompt words and calling a preset professional knowledge base; using a preset target large language model to parse the target policy data based on the target prompt words and the professional knowledge base to obtain corresponding interpretation data; and outputting the interpretation data. Furthermore, this application also relates to blockchain technology, allowing the interpretation data to be stored on the blockchain. This application can be applied to policy data parsing scenarios in the fintech field, effectively improving the efficiency of policy data parsing and processing.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology and can be applied to the financial technology field, particularly to data analysis methods, devices, computer equipment and storage media based on artificial intelligence. Background Technology

[0002] In the traditional insurance industry operation model, insurance companies are highly sensitive to policies issued by relevant authorities. Their ability to perceive, understand, and apply policies plays a decisive role in the future direction of insurance products, hence the extreme importance placed on policies by all insurance companies. However, the current method of policy analysis mainly relies on manual reading of large amounts of information from websites, then sifting through a massive amount of policy information to extract a small amount of content relevant to the insurance industry. This method not only consumes a large amount of manpower but also suffers from significant time lag, resulting in extremely low efficiency in policy analysis and processing.

[0003] Specifically, traditional policy analysis methods are limited to manual operation, making it difficult to quickly and comprehensively process large amounts of policy documents. Omissions or errors are prone to occur during information screening and extraction. This inefficient analysis model prevents insurance companies from grasping policy dynamics in a timely and accurate manner, thus affecting the development, adjustment, and marketing strategies of insurance products. For example, in property insurance within the financial insurance sector, when relevant authorities issue safety production policies targeting specific industries (such as manufacturing), traditional manual analysis methods may be too slow to promptly apply key information from the policies (such as new safety production standards and risk assessment requirements) to the design and pricing of property insurance products. If manufacturing enterprises need to increase their safety production investment due to new policies, and insurance companies fail to adjust property insurance product terms and rates in a timely manner, the products may fail to meet the actual needs of the enterprises, resulting in missed market opportunities and impacting the insurance companies' business development and market competitiveness.

[0004] Therefore, there is an urgent need to provide an efficient policy analysis method to improve the efficiency and accuracy of policy analysis, and to help insurance companies grasp policy directions in a timely manner and optimize insurance products and services. Summary of the Invention

[0005] The purpose of this application is to propose a data analysis method, apparatus, computer device, and storage medium based on artificial intelligence, so as to solve the technical problem of low processing efficiency in existing policy analysis methods.

[0006] Firstly, an artificial intelligence-based data parsing method is provided, including: Retrieve policy data from a pre-defined data source; The policy data is cleaned to obtain the corresponding first policy data; The first policy data is scored to obtain score data; Based on the scoring data, target policy data that meets the preset high-score criteria are selected from the first policy data; Retrieve preset target prompts and call up preset professional knowledge bases; Based on the target prompt words and the professional knowledge base, the target policy data is parsed and processed using a preset target large language model to obtain the corresponding interpretation data. The interpreted data is then processed for output.

[0007] Secondly, an artificial intelligence-based data analysis device is provided, including: The first acquisition module is used to acquire policy data from a preset data source; The cleaning module is used to clean the policy data to obtain the corresponding first policy data. The scoring module is used to score the first policy data to obtain scoring data. The filtering module is used to filter target policy data that meets preset high-score conditions from the first policy data based on the scoring data; The processing module is used to obtain preset target prompt words and call preset professional knowledge base; The parsing module is used to perform data parsing processing on the target policy data based on the target prompt words and the professional knowledge base, using a preset target large language model to obtain corresponding interpretation data; The output module is used to process the interpreted data.

[0008] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described artificial intelligence-based data parsing method.

[0009] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the aforementioned artificial intelligence-based data parsing method.

[0010] In the aforementioned scheme implemented by the AI-based data parsing method, apparatus, computer equipment, and storage medium, policy data is first obtained from a preset data source; then, the policy data is cleaned to obtain corresponding first policy data; subsequently, the first policy data is scored to obtain score data; and target policy data meeting preset high-score conditions is selected from the first policy data based on the score data; next, preset target prompt words are obtained, and a preset professional knowledge base is invoked; further, based on the target prompt words and the professional knowledge base, a preset target large language model is used to perform data parsing processing on the target policy data to obtain corresponding interpretation data; finally, the interpretation data is output. Unlike existing manual parsing methods, based on the above automated processing flow, this application cleans the policy data obtained from the data source to obtain first policy data, scores the first policy data to obtain score data, then selects target policy data meeting preset high-score conditions from the first policy data based on the score data, and then uses a preset target large language model to perform data parsing processing on the target policy data based on the obtained target prompt words and the invoked professional knowledge base to obtain interpretation data and output it. Thus, by combining target prompts, a professional knowledge base, and a target large language model, this application can efficiently and accurately complete the data parsing and processing of policy data, improving the efficiency of policy data parsing and processing, and ensuring the accuracy of the generated interpretation data. Attached Figure Description

[0011] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is an exemplary system architecture diagram to which this application can be applied; Figure 2 This is a flowchart of an embodiment of the artificial intelligence-based data parsing method according to this application; Figure 3 This is a schematic diagram of a structure of an embodiment of the artificial intelligence-based data parsing apparatus according to this application; Figure 4 This is a schematic diagram of the structure of one embodiment of the computer device according to this application. Detailed Implementation

[0013] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0014] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0015] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0016] like Figure 1 As shown, system architecture 100 may include terminal device 101, network 102, and server 103. Terminal device 101 may be a laptop 1011, tablet 1012, or mobile phone 1013. Network 102 is used as a medium to provide a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0017] Users can use terminal device 101 to interact with server 103 via network 102 to receive or send messages, etc. Various communication client applications can be installed on terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0018] Terminal device 101 can be various electronic devices with a display screen and support web browsing. In addition to laptops 1011, tablets 1012, or mobile phones 1013, terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer, and a desktop computer, etc.

[0019] Server 103 can be a server that provides various services, such as a backend server that provides support for the pages displayed on terminal device 101.

[0020] It should be noted that the AI-based data parsing method provided in this application is generally executed by a server / terminal device, and correspondingly, the AI-based data parsing device is generally located in the server / terminal device.

[0021] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0022] Continue to refer to Figure 2 The flowchart illustrates an embodiment of the AI-based data parsing method according to this application. The order of steps in the flowchart can be changed, and some steps can be omitted, depending on different needs. The AI-based data parsing method provided in this application can be applied to any scenario requiring data parsing, and therefore can be applied to products in these scenarios, such as data parsing products in the financial insurance field. The AI-based data parsing method includes the following steps: Step S201: Obtain policy data from a preset data source.

[0023] In this embodiment, the artificial intelligence-based data parsing method runs on an electronic device (e.g., Figure 1The server / terminal device shown can acquire policy data via wired or wireless connection. It should be noted that the aforementioned wireless connection methods include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra-wideband) connections, and other currently known or future-developed wireless connection methods. The implementing entity of this application is specifically a data parsing system, which can be simply referred to as the system. This application can be applied to policy data parsing scenarios in the financial and insurance fields. Specifically, it can obtain various policy data released by various government agencies by connecting with policy data sources externally sourced by insurance companies. These policy data sources are extensive, covering policy information released by government departments, national agencies, and other channels, forming the basis for subsequent screening and mining of policies related to the insurance industry.

[0024] Step S202: Clean the policy data to obtain the corresponding first policy data.

[0025] In this embodiment, the specific implementation process of cleaning the policy data to obtain the corresponding first policy data will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0026] Step S203: The first policy data is scored to obtain score data.

[0027] In this embodiment, the specific implementation process of scoring the first policy data to obtain the scoring data will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0028] Step S204: Based on the scoring data, select target policy data that meets the preset high-score conditions from the first policy data.

[0029] In this embodiment, meeting the preset high-score condition means that the score data is greater than a preset score threshold. Specifically, designated policy data whose score data is greater than the preset score threshold can be selected from all first policy data to serve as the target policy data meeting the high-score condition. Furthermore, the value of the aforementioned score threshold is not specifically limited and can be set according to actual business needs.

[0030] Step S205: Obtain preset target prompts and call preset professional knowledge base.

[0031] In this embodiment, the generation process of the aforementioned target prompts includes: designing a series of precise prompts based on the needs of policy interpretation, such as "Please analyze in detail the application scenarios of this policy in property insurance business, including the types of insurance products that may be involved, changes in policyholders, etc.", "Please assess the opportunities that this policy brings to property insurance business, such as market expansion, new business growth points, etc., and explain the reasons," and "Please point out the potential risks that this policy may bring to property insurance business, such as increased loss ratios, intensified competition, etc., and propose countermeasures." Alternatively, corresponding prompts can be set according to the goals and focus of the interpretation. Prompts can guide the model to interpret the policy from a specific perspective, such as prompting the model to analyze the application scenarios of the policy in the insurance industry, the potential business opportunities, and potential risk points. Prompts can also specify the format and length of the interpretation content to make the generated interpretation results more in line with the requirements.

[0032] The construction process of the aforementioned professional knowledge base includes: collecting and organizing professional knowledge bases provided by government and insurance experts, including relevant laws and regulations, business operation standards, and historical policy interpretation cases in the insurance industry. The professional knowledge base is then structured to facilitate rapid access and reference by the large-scale model when interpreting policies. This can be achieved by integrating the optimized target prompts and professional knowledge base with the target large-scale language model. By adjusting the model's parameters and configuration, the model can better combine prompts and professional knowledge for policy interpretation.

[0033] Step S206: Based on the target prompt words and the professional knowledge base, use a preset target large language model to perform data parsing processing on the target policy data to obtain the corresponding interpretation data.

[0034] In this embodiment, the data parsing process includes: Model understanding and analysis: Semantic understanding: The target large language model (hereinafter referred to as the model) utilizes the language knowledge and insurance domain knowledge learned during pre-training and fine-tuning to perform semantic understanding of the input policy text. The model analyzes keywords, phrases, and sentence structures in the policy text to understand the policy's intent, objectives, and main measures. For example, the model can identify requirements for insurance product innovation and strengthened regulation of the insurance market. Related knowledge retrieval: Based on understanding the policy text, the target large language model invokes its internally stored professional knowledge base and relevant knowledge learned during training to associate the policy content with the actual situation of the insurance industry. The model considers the impact of the policy on different types of insurance and different business processes, as well as the potential market reactions and industry changes. For example, when interpreting a text about health insurance policy, the model will consider knowledge such as the market size, competitive landscape, and consumer demand of health insurance, providing a more comprehensive perspective for subsequent interpretation.

[0035] Interpretation Content Generation: Application Scenario Analysis: Based on the target prompts, the model first analyzes the application scenarios of the policy in the insurance industry. The model will elaborate on how the policy is implemented in actual business, taking into account the characteristics and actual needs of insurance business. For example, for a policy encouraging insurance companies to develop agricultural insurance, the model will analyze how the policy will drive insurance companies to develop insurance products suitable for agricultural production, expand the rural insurance market, and improve the level of agricultural risk protection. Opportunity Assessment: Next, the target language model assesses the business opportunities that the policy may bring to the insurance industry. The model will analyze aspects such as market size expansion, product innovation, and customer base expansion, predicting the potential direction and potential growth of insurance business after the policy's implementation. For example, the policy may promote the rapid development of the health insurance market, providing more business opportunities for insurance companies. The model will analyze the size of this opportunity, its likelihood of realization, and the strategies that insurance companies should adopt. Risk Identification: The model will also identify potential risks that the policy may bring. Risks may include market risk, competition risk, and compliance risk. The model will analyze the causes of these risks, their potential impact, and how insurance companies should respond. For example, the policy may lead to increased competition in the insurance market. The model will indicate the impact of this risk on the market share and profits of insurance companies and propose corresponding countermeasures. Recommendation Generation: Finally, based on the analysis of opportunities and risks, the model generates corresponding recommendations. These recommendations should be targeted and actionable, helping insurance companies fully utilize policy opportunities and effectively address potential risks. For example, regarding the risk of intensified market competition, the model might recommend that insurance companies strengthen product innovation, improve service quality, and optimize marketing strategies.

[0036] The specific construction process of the aforementioned target large language model will be described in more detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0037] Step S207: Output the interpreted data.

[0038] In this embodiment, the specific implementation process of outputting the interpreted data described above will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0039] This application first obtains policy data from a preset data source; then cleans the policy data to obtain corresponding first policy data; next, it scores the first policy data to obtain score data; and based on the score data, it filters out target policy data that meets preset high-score criteria from the first policy data; subsequently, it obtains preset target prompt words and calls a preset professional knowledge base; further, based on the target prompt words and the professional knowledge base, it uses a preset target large language model to perform data parsing processing on the target policy data to obtain corresponding interpretation data; finally, it outputs the interpretation data. Unlike existing manual parsing methods, based on the above automated processing flow, this application obtains first policy data by cleaning the policy data obtained from the data source, scores the first policy data to obtain score data, then filters out target policy data that meets preset high-score criteria from the first policy data based on the score data, and then uses a preset target large language model to perform data parsing processing on the target policy data based on the obtained target prompt words and the called professional knowledge base to obtain interpretation data and output it. Thus, by combining target prompts, a professional knowledge base, and a target large language model, this application can efficiently and accurately complete the data parsing and processing of policy data, improving the efficiency of policy data parsing and processing, and ensuring the accuracy of the generated interpretation data.

[0040] In some alternative implementations, step S202 includes the following steps: Get the preset cleaning rules.

[0041] In this embodiment, the aforementioned cleaning rules can be cleaning rules formulated by insurance business experts based on the characteristics and needs of the insurance industry, used to identify irrelevant policy types (such as policies related to culture and education, sports events).

[0042] Based on the cleaning rules, the policy data is filtered to remove irrelevant data corresponding to the target industry, thereby obtaining the corresponding specified data.

[0043] In this embodiment, the acquired policy data can be screened one by one according to the constructed cleaning rules. Possible irrelevant policies can be identified by keyword filtering (such as "school construction" and "sports competition"), thereby identifying policies that are obviously unrelated to insurance (i.e., the target industry) to obtain the corresponding specified data.

[0044] The specified data is removed from the policy data to obtain the corresponding second policy data.

[0045] In this embodiment, the selected data can be removed from the above policy data, and the resulting second policy data can be used as the final first policy data.

[0046] The policy data is used as the first policy data.

[0047] In this embodiment, the first policy data obtained after cleaning can be backed up, and the cleaning process and results (number of policies removed, approximate situation of remaining policies) can be recorded to provide a basis for subsequent review and traceability.

[0048] This application obtains preset cleaning rules; then, based on these rules, it filters policy data for irrelevant data corresponding to the target industry to obtain corresponding specified data; subsequently, it removes the specified data from the policy data to obtain corresponding second policy data; and finally, it uses this policy data as the first policy data. Based on this process, this application obtains specified data by using cleaning rules to filter policy data for irrelevant data corresponding to the target industry, then removes the specified data from the policy data, and uses the obtained policy data as the required first policy data. This allows for the automatic and accurate removal of information irrelevant to the target industry from policy data, thereby narrowing the data scope and improving the efficiency and accuracy of subsequent data processing of the first policy data.

[0049] In some optional implementations of this embodiment, step S203 includes the following steps: Obtain the preset factor evaluation strategy.

[0050] In this embodiment, the above-mentioned factor assessment strategy refers to the assessment strategy corresponding to the three factors: the degree of correlation between policies and the insurance industry, the importance and influence of policies, and the risks brought by policies. The corresponding strategy content includes: 1. Scoring criteria for the degree of correlation between policies and the insurance industry. For policies directly related to insurance business: Policies are clearly formulated for insurance business, such as insurance product approval, insurance premium rate supervision, and insurance sales standards. These policies directly affect the operation and business of insurance companies and are scored higher, which can be set at 8-10 points. For policies indirectly affecting insurance business: Policies that are not directly targeted at insurance business but have a certain impact on the insurance industry, such as macroeconomic policies, tax policies, and social security policies. The impact of these policies is relatively weak, and the score can be set at 5-7 points. For policies with low correlation: Policies with low correlation to the insurance industry and little impact on insurance business, such as some industrial policies and environmental protection policies unrelated to insurance business. These policies are scored lower and can be set at 1-4 points.

[0051] 2. Policy Importance and Impact Scoring Criteria. **High Market Attention:** Policies that have garnered widespread attention and discussion in the market, such as major policy reforms and adjustments to industry regulatory policies. These policies typically have a significant impact on the insurance market landscape and are rated higher, around 8-10 points. **Some Impact:** Policies with a relatively limited impact on the overall insurance market, but may have some influence on certain specific areas or market segments, such as regional insurance policies or adjustments to policies for specific insurance products. These policies are rated 5-7 points. **Small Impact:** Policies with a small scope of impact and little influence on the overall structure and business development of the insurance market, such as some localized industry regulations and technical standards. These policies are rated lower, around 1-4 points.

[0052] 3. Risk Scoring Criteria for Policy-Related Risks. High Risk: Policy changes may significantly increase the risk of insurance business. For example, a reduction in insurance premiums may lead to a decrease in insurance company profits and an increase in the insurance loss ratio. These policies are rated higher, with a score of 8-10. Medium Risk: Policy changes have some impact on the risk of insurance business, but the degree of risk is relatively controllable. For example, policies on insurance product innovation may bring certain market and compliance risks. These policies are rated with a score of 5-7. Low Risk: Policy changes have a small impact on the risk of insurance business, or the risk can be controlled through effective risk management measures, such as some routine adjustments to industry regulations. These policies are rated lower, with a score of 1-4.

[0053] The first policy data is processed by factor evaluation based on the aforementioned factor evaluation strategy to obtain the corresponding factor evaluation results.

[0054] In this embodiment, the first policy data can be processed for factor evaluation based on the strategy content of the above factor evaluation strategy to obtain the corresponding factor evaluation results.

[0055] The evaluation results of the factors are scored and calculated based on preset scoring rules to obtain the calculation results corresponding to the first policy data.

[0056] In this embodiment, weights corresponding to various factors can be set according to actual business needs. For example, the relevance of the policy to the insurance industry accounts for 40% of the weight, the importance and influence of the policy accounts for 30% of the weight, and the risks brought by the policy account for 30% of the weight. Then, based on the factor evaluation results of each factor, a weighted sum is performed according to the weights to calculate the comprehensive score of the policy, i.e., the calculation result mentioned above.

[0057] The calculation results are used as the scoring data for the first policy data.

[0058] In this embodiment, controversial scores can also be reassessed and adjusted to ensure accuracy and reasonableness. The final high-scoring policy data is then incorporated into the large-scale model for policy interpretation.

[0059] Assigning scores is a crucial method for further filtering and prioritizing policy data. By comprehensively considering factors such as the policy's relevance to the insurance industry, its importance and impact, and potential risks, a comprehensive score is assigned to each policy data point. This allows for prioritization of policy data based on scores, identifying policies with a significant impact on the insurance industry and strong relevance to insurance business, facilitating subsequent focused interpretation and analysis, and improving the efficiency of policy resource utilization.

[0060] This application obtains a preset factor assessment strategy; then, based on the factor assessment strategy, it performs factor assessment processing on the first policy data to obtain the corresponding factor assessment results; subsequently, it performs score calculation processing on the factor assessment results based on preset scoring rules to obtain the calculation results corresponding to the first policy data; and finally, it uses the calculation results as the score data for the first policy data. Based on the above processing flow, this application obtains factor assessment results by performing factor assessment processing on the first policy data based on the use of a factor assessment strategy, and then performs score calculation processing on the factor assessment results based on the use of scoring rules, and uses the obtained calculation results corresponding to the first policy data as the corresponding score data. This enables automatic and intelligent completion of the scoring processing of the first policy data, improves the processing efficiency of the scoring process, and ensures the accuracy of the obtained score data.

[0061] In some alternative implementations, prior to step S206, the electronic device may also perform the following steps: Acquire pre-collected multi-source data.

[0062] In this embodiment, the process of generating the aforementioned multi-source data includes: Multi-source data aggregation: Policy text collection: A wide range of policy texts related to the insurance industry are collected from government official websites, authoritative policy release platforms, and insurance industry regulatory agency websites. These policies cover macro-level policy guidance, such as strategic planning for the overall development of the insurance industry; they also include policies for specific business areas, such as regulations and requirements for specific insurance types like health insurance, auto insurance, and agricultural insurance. Insurance business data collection: Business data from insurance companies is collected, including sales data, claims data, and customer information for different insurance types. This data helps the model understand the actual operation and business model of the insurance industry, providing practical evidence for policy interpretation. Industry research report collection: Research reports and market analysis reports on the insurance industry are obtained. These reports typically contain in-depth analysis of industry trends, competitive landscape, and market demand, providing the model with both macro and micro-level industry perspectives.

[0063] Data Cleaning and Labeling: Data Cleaning: Cleaning the collected multi-source data to remove duplicates, errors, and incomplete data. For example, correcting formatting errors and typos in policy texts; handling outliers and missing values ​​in business data to ensure data quality and consistency. Data Labeling: Labeling policy texts to highlight key policy information, such as policy objectives, scope of application, implementation time, and main measures. Simultaneously, labeling insurance business data and industry research reports to highlight policy-related business scenarios and market reactions. Labeling can be performed by professional policy and insurance experts and insurance business personnel to ensure accuracy and professionalism. Alternatively, a semi-automatic labeling approach can be used. For policies that clearly meet labeling characteristics, semi-automatic labeling can be achieved by setting rules to improve labeling efficiency.

[0064] Invoke the preset initial large language model.

[0065] In this embodiment, a suitable natural language processing model architecture can be selected based on the needs of the policy interpretation task, such as a model based on the Transformer architecture. The Transformer architecture has powerful language understanding and generation capabilities, can handle long texts and complex semantic relationships, and is suitable for in-depth interpretation of policy texts. Alternatively, the GPT model, Deepseek model, etc., can also be used as the initial large language model mentioned above.

[0066] The initial large language model is pre-trained based on the multi-source data to obtain the corresponding first large language model.

[0067] In this embodiment, the initial large language model can be pre-trained using the aforementioned multi-source model. Through pre-training, the initial large language model learns the basic rules and semantic representations of language, laying the foundation for subsequent fine-tuning tasks. Specifically, during pre-training, professional knowledge corpora from the insurance field are introduced, such as insurance regulations, insurance clauses, and insurance industry terminology. Further training on these insurance-related corpora allows the initial large language model to gain a deeper understanding of the insurance industry, improving its performance in insurance policy interpretation tasks and resulting in a pre-trained first large language model.

[0068] The first language model is fine-tuned and trained based on a pre-defined adversarial training module to obtain the corresponding second language model.

[0069] In this embodiment, the fine-tuning training process includes: **Cue Word Engineering Optimization:** Designing a series of carefully crafted prompt words to guide the model in generating text that meets policy interpretation requirements. Prompt words can include descriptions of the policy background, the angle and focus of interpretation, and the desired output format. For example, a prompt word could be, "Please provide a detailed interpretation of the following policy from the perspective of insurance industry applications, opportunities, and potential risks: [Policy Text]". Continuously optimizing the prompt words improves the quality and relevance of the model-generated content. **Introduction of an Adversarial Training Module:** To reduce model illusions and improve the effectiveness of the model's parsing, an adversarial training module is introduced. Adversarial training generates adversarial samples, allowing the model to learn how to distinguish between real and valid information and false or erroneous information during training. For example, adversarial samples similar to real policy texts but containing erroneous information can be generated, allowing the model to identify and correct them, thereby enhancing the model's robustness and accuracy. **Expert Knowledge Base Integration:** Integrating the professional knowledge base provided by government and insurance experts with the model. The professional knowledge base includes expert experience, case analyses, and industry standards in the insurance industry. During fine-tuning, the model can refer to the information in the professional knowledge base to ensure that the interpreted content conforms to industry realities and professional standards. For example, when interpreting policy changes involving insurance clauses, models can refer to case studies and impact analyses of similar clause changes in the professional knowledge base to improve the accuracy of the interpretation.

[0070] The first language model can be fine-tuned based on the above-mentioned fine-tuning training steps to obtain the fine-tuned second language model.

[0071] The second language model is optimized based on a preset optimization strategy to obtain the corresponding third language model.

[0072] In this embodiment, the optimization strategy includes: **Evaluation Indicator Setting:** A series of evaluation indicators are set to measure the effectiveness of the model's interpretation, such as accuracy, completeness, logic, and practicality. Accuracy indicators measure the degree to which the model's interpretation conforms to the original policy text and actual situation; completeness indicators measure whether the model comprehensively interprets all aspects of the policy; logic indicators measure whether the logic of the model-generated content is clear and coherent; and practicality indicators measure the value of the model's interpretation for practical application in the insurance industry. **Manual Evaluation and Feedback:** Government and insurance experts and insurance professionals conduct manual evaluations of the model's interpretation. Evaluators score and evaluate the interpretation based on the set evaluation indicators and provide specific improvement suggestions. For example, evaluators may point out misunderstandings, omissions, or logical errors in the interpretation and provide correct interpretations and suggestions. **Model Optimization and Iteration:** Based on the feedback from the manual evaluation, the model is optimized and iterated. This includes adjusting model parameters, optimizing prompt word engineering, and updating the professional knowledge base to continuously improve model performance and interpretation quality. Through multiple iterations, the effectiveness of the model's parsed content is stabilized at over 80%.

[0073] Specifically, the second largest language model can be optimized based on the above optimization strategy, and the resulting third largest language model can be used as the final target large language model.

[0074] The third major language model is used as the target major language model.

[0075] This application obtains pre-collected multi-source data; calls a preset initial large language model; then pre-trains the initial large language model based on the multi-source data to obtain a corresponding first large language model; subsequently, fine-tunes the first large language model based on a preset adversarial training module to obtain a corresponding second large language model; subsequently, optimizes the second large language model based on a preset optimization strategy to obtain a corresponding third large language model; finally, the third large language model is used as the target large language model. Based on the above processing flow, this application obtains a first large language model by pre-training the initial large language model using pre-collected multi-source data, then fine-tunes the first large language model based on the use of an adversarial training module to obtain a second large language model, then optimizes the second large language model based on an optimization strategy, and uses the resulting third large language model as the final target large language model. This allows for efficient and accurate model construction of the target large language model while ensuring the model performance of the obtained target large language model.

[0076] In some alternative implementations, step S207 includes the following steps: Get the preset organization strategy.

[0077] In this embodiment, the strategy of organizing the data includes: organizing and arranging the interpretation data generated by the target large language model to form a well-structured and logically coherent interpretation report. The interpretation report can be organized into chapters such as application scenario analysis, opportunity assessment, risk identification, and response suggestions, with each chapter further subdivided into specific key points and content.

[0078] Based on the aforementioned organization strategy, the interpreted data is organized and arranged to obtain a corresponding interpretation report.

[0079] In this embodiment, the interpreted data can be organized and arranged based on the strategy content of the above-mentioned organization strategy to obtain a processed interpretation report.

[0080] The interpretation report is then standardized to obtain the corresponding target interpretation report.

[0081] In this embodiment, the standardization process includes: formatting and beautifying the interpretation report to ensure readability and professionalism. Specifically, elements such as titles, paragraph indentation, and bullet points can be added to make the report structure clearer; and charts, case studies, and other supplementary explanations can be inserted to enhance the report's visualization and persuasiveness. The final output is a high-quality policy interpretation report, i.e., a target interpretation report, providing comprehensive reference for insurance companies' decision-making.

[0082] The target interpretation report is then output and processed.

[0083] In this embodiment, the output processing of the target interpretation report can be completed by sending the generated target interpretation report to the corresponding target organization.

[0084] This application obtains a preset data processing strategy; then, based on the strategy, it processes and arranges the interpreted data to obtain a corresponding interpretation report; subsequently, it standardizes the interpretation report to obtain a target interpretation report; and finally, it outputs the target interpretation report. Based on this process, this application effectively improves the accuracy and practicality of the generated target interpretation report by processing and arranging the interpreted data using a data processing strategy to obtain an interpretation report, and then standardizing the interpretation report to obtain and output the target interpretation report.

[0085] In some optional implementations of this embodiment, after step S203, the electronic device may further perform the following steps: Obtain the preset labeling strategy.

[0086] In this embodiment, the labeling system, i.e., the labeling strategy, can be discussed and agreed upon by insurance business experts and label designers. Labels such as "Financial Sector Policies" and "Agricultural Sector Policies" are designed based on the relevant fields; labels such as "Tax Policies," "Regulatory Policies," and "Subsidy Policies" are designed based on the policy type; and for insurance-related policies, labels such as "Property Insurance Related Policies," "Life Insurance Related Policies," and "Agricultural Insurance Related Policies" are further refined. Furthermore, labeling standards and procedures are established, clarifying the meaning and scope of use of each label. For example, for the "Property Insurance Related Policies" label, it is stipulated that it can only be labeled when the policy content directly or indirectly affects property insurance business (such as affecting enterprises' willingness to insure their property, changing the method of calculating property insurance rates, etc.).

[0087] The target policy data is labeled based on the labeling strategy to obtain the corresponding third policy data.

[0088] In this embodiment, the cleaned target policy data can be labeled based on the labeling specifications and processes in the above-mentioned labeling strategy. A combination of manual and semi-automatic labeling methods can be used. For policies that clearly meet the labeling characteristics, semi-automatic labeling can be achieved by setting rules, thereby improving labeling efficiency.

[0089] The third policy data is stored and processed.

[0090] In this embodiment, the storage method for the aforementioned third policy data is not specifically limited and can be determined according to actual business needs. For example, local database storage, cloud storage, blockchain storage, etc. can be used.

[0091] This application obtains a preset tagging strategy; then, based on the tagging strategy, it performs tagging processing on the target policy data to obtain corresponding third policy data; subsequently, it stores the third policy data. Based on the above processing flow, this application obtains third policy data by tagging target policy data using a tagging strategy, and then stores the third policy data, thereby enabling the classification and management of third policy data, facilitating subsequent querying and analysis. By classifying the third policy data according to different tags, users can quickly find the policy data they need, which is beneficial to the readability and usability of policy data, and provides convenience for subsequent policy interpretation and application.

[0092] In some optional implementations of this embodiment, after step S206, the electronic device may further perform the following steps: The target policy data and the interpretation data are processed to obtain the corresponding integrated data.

[0093] In this embodiment, the generated policy interpretation reports and high-scoring policies can be sorted and packaged to obtain the sorted integrated data, or policy package.

[0094] Invoke the preset distribution channel.

[0095] In this embodiment, the selection of the aforementioned distribution channels is not specifically limited and can be determined based on actual business needs. For example, a suitable system platform can be selected as the channel for policy dissemination. This platform should be secure, stable, and easy to use. Functional testing and optimization of the platform should be conducted to ensure smooth uploading and downloading of policies and interpretations.

[0096] Identify the target institutions to be issued the documents.

[0097] In this embodiment, the aforementioned target organization may refer to branch offices in various regions.

[0098] Based on the aforementioned distribution channel, the integrated data is sent to the target organization.

[0099] In this embodiment, the generated integrated data, i.e., the policy package, can be distributed to the corresponding institutions through selected distribution channels according to their organizational structure and hierarchical relationships. During the distribution process, information such as the distribution time and the receiving institution is recorded for subsequent tracking and querying. Simultaneously, the heads of each target institution can be notified via SMS, email, or other means to remind them to check for the policy content in a timely manner.

[0100] To ensure timely feedback from institutions regarding policies, system tracking tasks are generated simultaneously with policy issuance. These tasks clearly define the work content and timelines that institutions need to complete, such as requiring them to provide feedback within a certain timeframe on the impact of the policy on their projects and potential business opportunities. In addition to system tracking tasks, email tracking and notifications further strengthen the monitoring of institutional feedback. Regular reminders are sent to institutions via email to urge them to complete tasks on time and provide timely feedback, ensuring the effective implementation and execution of policies across all institutions.

[0101] In addition, the system also has functions for institutional feedback and business opportunity generation, including: 1) The process of institutions providing feedback based on policies: After receiving the policies and interpretations, each institution organizes relevant business personnel for discussion and analysis. Business personnel, combining their own business projects and market conditions, delve into the potential business opportunities brought by the policies. For example, regarding a policy on corporate tax incentives, business personnel can analyze whether the policy will encourage companies to increase their property insurance coverage, thereby uncovering opportunities to increase premium income. Institutions, according to the tracking task requirements on the system platform, organize and provide feedback on the discovered business opportunities and expected outputs. Feedback includes a detailed description of the business opportunity, the expected business volume (such as new premium income, number of new policies, etc.), and the expected implementation time. Institution leaders review and summarize the feedback from business personnel to ensure the accuracy and completeness of the feedback information. 2) The process of supplementing output information and excellent case studies after output: After producing relevant business results, institutions promptly supplement the output information to the system platform. Output information includes actual business volume (such as actual new premium income, actual new customer numbers, etc.), problems encountered during business operations, and solutions. Simultaneously, institutions should identify and compile outstanding cases generated during policy application. Outstanding cases should be representative, replicable, and referential; for example, cases where an institution successfully expanded into new markets or increased its customer base through clever policy application. Detailed information on outstanding cases (including case background, specific methods of policy application, and achieved results) should be fed back to the system platform. 3) The process of sending the information to institutions for learning after headquarters review: Headquarters organizes relevant personnel to review the output information and outstanding cases submitted by institutions. The review includes the authenticity, accuracy, and completeness of the information, as well as the typicality and referential value of the cases. For approved information and cases, headquarters will organize and compile them into learning materials. Learning materials may include case analysis reports, experience-sharing documents, etc. The learning materials will be sent to institutions through the system platform. Simultaneously, online or offline training activities can be organized to interpret and explain the learning materials, ensuring that institutions can fully understand and absorb the experience and knowledge contained therein.

[0102] This application processes the target policy data and the interpretation data to obtain integrated data; then it calls a preset distribution channel; next, it identifies the target institutions to which the data will be distributed; and finally, based on the distribution channel, it sends the integrated data to the target institutions. Based on this process, this application obtains integrated data by processing the target policy data and interpretation data, and then sends the integrated data to the target institutions using the distribution channel, ensuring that the target institutions can understand policy developments in a timely manner and adjust their business accordingly. Furthermore, the use of the distribution channel improves the efficiency and accuracy of policy information transmission, avoiding the information loss or delays that may occur with traditional methods.

[0103] In some alternative implementations, the user information obtained is subject to user consent and complies with relevant laws and policies.

[0104] Furthermore, any software tools or components not belonging to our company that appear in the embodiments of this application are merely illustrative examples and do not represent actual use.

[0105] In addition, the system also has the function of building a political security map system, including: 1. The process of integrating multi-perspective and multi-dimensional data: By defining the perspectives and dimensions for data integration, including regional dimensions (such as different provinces and cities), business type dimensions (such as property insurance and life insurance), and policy type dimensions (such as tax policies and regulatory policies), data relevant to these perspectives and dimensions is extracted from the system platform and other relevant data sources. This includes data such as the number of policies distributed across different regions, the degree to which different business types are affected by policies, and the effectiveness of policy implementation. The extracted data is then cleaned and preprocessed to ensure accuracy and consistency. This includes standardizing data formats and handling missing and outlier values. Finally, the processed data is integrated according to different perspectives and dimensions to establish a data warehouse or data mart, providing data support for subsequent visualization.

[0106] The integration of multi-perspective and multi-dimensional data aims to provide a comprehensive and in-depth understanding of the application of policies and market dynamics within the insurance industry. By analyzing data from different angles and dimensions, it is possible to identify the differences and impacts of policies across different regions and business types, providing a basis for headquarters to formulate targeted business strategies. The establishment of a data warehouse or data mart facilitates data storage and management, improving data availability.

[0107] 2. The implementation process is presented intuitively and visually: By selecting appropriate visualization tools, such as Tableau and Power BI, and designing the types and layouts of visualization charts based on the characteristics of the integrated data and display requirements, the following steps can be taken. For example, maps can be used to display the policy distribution in different regions, bar charts can be used to show the degree to which different business types are affected by policies, and line charts can be used to show the changes in the policy implementation effect over time. The visualization charts are then beautified and optimized by adding elements such as titles, labels, and legends to improve readability and aesthetics. Simultaneously, the interactivity of the charts is ensured, allowing users to obtain more detailed information through clicks, filtering, and other operations. Subsequently, the visualization charts will be integrated into a unified government security map system platform for convenient centralized viewing and analysis by users. The platform will provide search, filtering, and sorting functions, enabling users to quickly find the chart information they need based on their requirements.

[0108] In particular, intuitive visualization presents the integrated data to managers in a more intuitive and easy-to-understand way. Through visual charts, managers can gain a clear understanding of market dynamics and the business situation of each organization at a glance.

[0109] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0110] It should be emphasized that, to further ensure the privacy and security of the aforementioned interpreted data, the interpreted data can also be stored in a node of a blockchain.

[0111] The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0112] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0113] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware through computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0114] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0115] Further reference Figure 3 As a response to the above Figure 2 To implement the method shown, this application provides an embodiment of an artificial intelligence-based data parsing device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0116] like Figure 3 As shown, the artificial intelligence-based data analysis device 300 described in this embodiment includes: a first acquisition module 301, a cleaning module 302, a scoring module 303, a filtering module 304, a processing module 305, an analysis module 306, and an output module 307. Wherein: The first acquisition module 301 is used to acquire policy data from a preset data source; The cleaning module 302 is used to clean the policy data to obtain the corresponding first policy data; The scoring module 303 is used to score the first policy data to obtain scoring data; The filtering module 304 is used to filter target policy data that meets preset high-score conditions from the first policy data based on the scoring data; The processing module 305 is used to obtain preset target prompt words and call preset professional knowledge base; The parsing module 306 is used to perform data parsing processing on the target policy data based on the target prompt words and the professional knowledge base, using a preset target large language model to obtain corresponding interpretation data; The output module 307 is used to output and process the interpreted data.

[0117] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data parsing method in the aforementioned implementation method, and will not be repeated here.

[0118] In some optional implementations of this embodiment, the cleaning module 302 includes: The first acquisition submodule is used to acquire preset cleaning rules; The filtering submodule is used to filter irrelevant data from the policy data based on the cleaning rules to obtain the corresponding specified data. The elimination submodule is used to remove the specified data from the policy data to obtain the corresponding second policy data. The first determining submodule is used to use the policy data as the first policy data.

[0119] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data parsing method in the aforementioned implementation method, and will not be repeated here.

[0120] In some optional implementations of this embodiment, the scoring module 303 includes: The second acquisition submodule is used to acquire preset factor evaluation strategies; The evaluation submodule is used to perform factor evaluation processing on the first policy data based on the factor evaluation strategy to obtain the corresponding factor evaluation results; The calculation submodule is used to perform scoring calculation on the evaluation results of the factors based on preset scoring rules, and obtain the calculation results corresponding to the first policy data; The second determining submodule is used to use the calculation result as the scoring data of the first policy data.

[0121] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data parsing method in the aforementioned implementation method, and will not be repeated here.

[0122] In some optional implementations of this embodiment, the artificial intelligence-based data parsing device further includes: The second acquisition module is used to acquire pre-collected multi-source data; The first calling module is used to call the preset initial large language model; The pre-training module is used to pre-train the initial large language model based on the multi-source data to obtain the corresponding first large language model. The fine-tuning module is used to fine-tune the first large language model based on the preset adversarial training module to obtain the corresponding second large language model. The optimization module is used to optimize the second language model based on a preset optimization strategy to obtain the corresponding third language model. The first determining module is used to select the third large language model as the target large language model.

[0123] In some optional implementations of this embodiment, the output module 307 includes: The third acquisition submodule is used to acquire preset sorting strategies; The first processing submodule is used to organize and arrange the interpretation data based on the organization strategy to obtain the corresponding interpretation report; The second processing submodule is used to standardize the interpretation report to obtain the corresponding target interpretation report; The output submodule is used to process the output of the target interpretation report.

[0124] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data parsing method in the aforementioned implementation method, and will not be repeated here. In some optional implementations of this embodiment, the artificial intelligence-based data parsing device further includes: The third acquisition module is used to acquire the preset marking strategy; The labeling module is used to perform labeling processing on the target policy data based on the labeling strategy to obtain the corresponding third policy data; The storage module is used to store and process the third policy data.

[0125] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data parsing method in the aforementioned implementation method, and will not be repeated here.

[0126] In some optional implementations of this embodiment, the artificial intelligence-based data parsing device further includes: The processing module is used to process the target policy data and the interpretation data to obtain the corresponding integrated data; The second calling module is used to call the preset distribution channels; The second determination module is used to determine the target organization to be issued; The sending module is used to send the integrated data to the target organization based on the distribution channel.

[0127] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data parsing method in the aforementioned implementation method, and will not be repeated here. To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0128] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected via a system bus. It should be noted that only the computer device 4 with components 41-43 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0129] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0130] The memory 41 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 41 may be an internal storage unit of the computer device 4, such as the hard disk or memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 4. Of course, the memory 41 may also include both the internal storage unit and its external storage device of the computer device 4. In this embodiment, the memory 41 is typically used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions based on artificial intelligence data parsing methods. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or will be output.

[0131] In some embodiments, the processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 42 is typically used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to execute computer-readable instructions stored in the memory 41 or to process data, for example, to execute computer-readable instructions of the artificial intelligence-based data parsing method.

[0132] The network interface 43 may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 4 and other electronic devices.

[0133] Compared with the prior art, the embodiments of this application have the following beneficial effects: In this embodiment, the application cleanses policy data obtained from a data source to obtain first policy data, scores the first policy data to obtain score data, and then filters target policy data that meets preset high-score criteria from the first policy data based on the score data. Furthermore, based on the obtained target prompts and the accessed professional knowledge base, a preset target large language model is used to perform data parsing processing on the target policy data to obtain interpreted data, which is then output. Thus, by combining the use of target prompts, a professional knowledge base, and a target large language model, this application can efficiently and accurately complete the data parsing processing of policy data, improving the efficiency of policy data parsing and ensuring the accuracy of the generated interpreted data.

[0134] This application also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the artificial intelligence-based data parsing method described above.

[0135] Compared with the prior art, the embodiments of this application have the following main advantages: In this embodiment, the application cleanses policy data obtained from a data source to obtain first policy data, scores the first policy data to obtain score data, and then filters target policy data that meets preset high-score criteria from the first policy data based on the score data. Furthermore, based on the obtained target prompts and the accessed professional knowledge base, a preset target large language model is used to perform data parsing processing on the target policy data to obtain interpreted data, which is then output. Thus, by combining the use of target prompts, a professional knowledge base, and a target large language model, this application can efficiently and accurately complete the data parsing processing of policy data, improving the efficiency of policy data parsing and ensuring the accuracy of the generated interpreted data.

[0136] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0137] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

Claims

1. A data parsing method based on artificial intelligence, characterized in that, Includes the following steps: Retrieve policy data from a pre-defined data source; The policy data is cleaned to obtain the corresponding first policy data; The first policy data is scored to obtain score data; Based on the scoring data, target policy data that meets the preset high-score criteria are selected from the first policy data; Retrieve preset target prompts and call up preset professional knowledge bases; Based on the target prompt words and the professional knowledge base, the target policy data is parsed and processed using a preset target large language model to obtain the corresponding interpretation data. The interpreted data is then processed for output.

2. The data parsing method based on artificial intelligence according to claim 1, characterized in that, The step of cleaning the policy data to obtain the corresponding first policy data specifically includes: Obtain the preset cleaning rules; Based on the cleaning rules, the policy data is filtered to remove irrelevant data corresponding to the target industry, thereby obtaining the corresponding specified data. The specified data is removed from the policy data to obtain the corresponding second policy data; The policy data is used as the first policy data.

3. The data parsing method based on artificial intelligence according to claim 1, characterized in that, The step of scoring the first policy data to obtain the score data specifically includes: Obtain the preset factor evaluation strategy; Based on the aforementioned factor assessment strategy, the first policy data is subjected to factor assessment processing to obtain the corresponding factor assessment results; The evaluation results of the factors are scored and calculated based on the preset scoring rules to obtain the calculation results corresponding to the first policy data; The calculation results are used as the scoring data for the first policy data.

4. The data parsing method based on artificial intelligence according to claim 1, characterized in that, Before the step of performing data parsing and processing on the target policy data using a preset target large language model based on the target prompt words and the professional knowledge base to obtain the corresponding interpretation data, the method further includes: Acquire pre-collected multi-source data; Invoke the preset initial large language model; Based on the multi-source data, the initial large language model is pre-trained to obtain the corresponding first large language model. The first language model is fine-tuned and trained based on a preset adversarial training module to obtain the corresponding second language model. The second language model is optimized based on a preset optimization strategy to obtain the corresponding third language model. The third major language model is used as the target major language model.

5. The data parsing method based on artificial intelligence according to claim 1, characterized in that, The step of outputting the interpreted data specifically includes: Obtain the preset organization strategy; Based on the aforementioned organization strategy, the interpreted data is organized and arranged to obtain a corresponding interpretation report; The interpretation report is standardized to obtain the corresponding target interpretation report; The target interpretation report is then output and processed.

6. The data parsing method based on artificial intelligence according to claim 1, characterized in that, After the step of filtering target policy data that meets preset high-score criteria from the first policy data based on the scoring data, the method further includes: Obtain the preset marking strategy; The target policy data is labeled based on the labeling strategy to obtain the corresponding third policy data. The third policy data is stored and processed.

7. The data parsing method based on artificial intelligence according to claim 1, characterized in that, After the step of parsing and processing the target policy data using a preset target large language model based on the target prompt words and the professional knowledge base to obtain the corresponding interpretation data, the method further includes: The target policy data and the interpretation data are processed to obtain the corresponding integrated data; Invoke the preset distribution channel; Identify the target institutions to be issued the documents; Based on the aforementioned distribution channel, the integrated data is sent to the target organization.

8. A data analysis device based on artificial intelligence, characterized in that, include: The first acquisition module is used to acquire policy data from a preset data source; The cleaning module is used to clean the policy data to obtain the corresponding first policy data. The scoring module is used to score the first policy data to obtain scoring data. The filtering module is used to filter target policy data that meets preset high-score conditions from the first policy data based on the scoring data; The processing module is used to obtain preset target prompt words and call preset professional knowledge base; The parsing module is used to perform data parsing processing on the target policy data based on the target prompt words and the professional knowledge base, using a preset target large language model to obtain corresponding interpretation data; The output module is used to process the interpreted data.

9. A computer device, characterized in that, The system includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the data parsing method based on artificial intelligence as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the data parsing method based on artificial intelligence as described in any one of claims 1 to 7.

Images