A policy interpretation method based on deep learning and an intelligent policy interpretation system
Through deep learning technology, data is collected and analyzed in multiple aspects in the policy interpretation system, a reasonableness index for policy interpretation is established, which solves the problems of insufficient comprehensive data collection and inaccurate analysis results in the existing technology, and improves the accuracy and rationality of policy interpretation. sex.
Patent Information
- Application Number
- CN202411172941.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-26
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-08-26
AI Technical Summary
The existing policy interpretation system is difficult to collect and analyze from multiple angles when collecting and analyzing data, resulting in inaccurate analysis results and the singularity of the data affects the rationality of the interpretation.
Using a policy interpretation method based on deep learning, model quality data, model training data, keyword data, semantic similarity data and adaptability data are collected in multiple aspects through policy data acquisition steps, and multi-angle analysis is carried out through corresponding mathematical models to calculate Extract the characteristic values of policy data, establish a policy interpretation rationality index, and provide real-time feedback on reasonableness judgments.
It improves the accuracy and rational judgment of the analysis results, reduces the adverse impact of data singularity, and ensures the scientificity and accuracy of policy interpretation.
Smart Images

Figure CN118673932B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of policy interpretation, and more specifically, to a policy interpretation method and an intelligent policy interpretation system based on deep learning. Background Art
[0002] With the rapid development of the economy, policy changes are frequent and complex. Traditional methods are difficult to fully capture the subtleties of policy changes. Through deep learning, we conduct in-depth mining and analysis of policy texts, aiming to construct a policy floating index, display policy change trends in real time, and provide scientific basis and decision-making support for policy makers and all sectors of society.
[0003] The deep learning policy interpretation method integrates advanced artificial intelligence technologies, especially natural language processing and deep learning algorithms. This method converts policy texts into a machine-understandable format and uses deep learning models to perform semantic analysis, sentiment recognition, and topic clustering on policy content. On this basis, the system can automatically extract the key terms, implementation goals, scope of influence, and potential impact of the policy, and generate a detailed interpretation report. In addition, by comparing historical policy data, the system can also analyze the trends and patterns of policy changes, and provide accurate and comprehensive policy interpretation services for policymakers, enterprises, and the public. This method not only improves the efficiency and accuracy of policy interpretation, but also promotes the deep integration of policy research and application.
[0004] However, it still has some shortcomings in actual use. For example, when collecting and analyzing data, it is difficult to collect and analyze data from multiple angles, which leads to inaccurate analysis results and the singleness of the collected data, thus affecting the rationality of the entire policy interpretation.
[0005] Therefore, there is an urgent need to provide a policy interpretation method and intelligent policy interpretation system based on deep learning to solve the problems of incomplete data collection and inaccurate data analysis results in the existing policy interpretation system. Summary of the invention
[0006] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a policy interpretation method and an intelligent policy interpretation system based on deep learning, and solves the problems raised in the above-mentioned background technology through the following scheme.
[0007] To achieve the above objectives, the present invention provides the following technical solutions: a policy interpretation method based on deep learning and an intelligent policy interpretation system, comprising:
[0008] Step P1: Policy data collection step: used to collect policy model quality data, model training data, keyword data, semantic similarity data and adaptability data, and transmit the collected data to the policy data analysis step;
[0009] Step P2: Policy data analysis step: including model quality data analysis unit, model training data analysis unit, keyword data analysis unit, semantic similarity data analysis unit and adaptability data analysis unit, establishing corresponding mathematical models through each analysis unit, importing the data transmitted in the policy data collection step into the corresponding mathematical model, calculating the model quality characteristic value, model training efficiency characteristic value, keyword rationality characteristic value, semantic similarity characteristic value and adaptability characteristic value of the policy data, and transmitting them to the policy interpretation data analysis step;
[0010] Step P3: Policy interpretation data analysis step: including a policy interpretation data analysis unit, which is used to analyze the data transmitted in the policy data analysis step and transmit the analysis results to the final evaluation step;
[0011] Step P4: Final evaluation step: used to establish a preset value of the policy interpretation rationality index, judge the policy rationality index value according to the preset value of the policy interpretation rationality index, and transmit the judgment result to the real-time feedback step;
[0012] Step P5: Real-time feedback step: used to receive the judgment result transmitted by the final evaluation step. When the policy interpretation is reasonable, a normal signal is sent; when the policy interpretation is unreasonable, an early warning signal is sent to the terminal management personnel.
[0013] Preferably, the model quality data includes the F1 score FM, the total text entropy value Te and the semantic vector dimension Sd; the model training data includes the policy text processing speed Ts, the number of standard entities Ae and the text reasoning time Rt; the keyword quality data includes the keyword coverage Kr, the keyword extraction speed Ks and the number of keywords Kn; the semantic similarity data includes the similarity threshold St, the similar policy document retrieval rate Dr and the retrieval response time Jt; the adaptability data includes the policy update frequency Uf, the policy abolition rate Ur and the domain migration loss rate Dm.
[0014] Preferably, the model quality data analysis unit is used to establish a model quality data calculation model, import the model quality data in the policy data collection step into the model quality data calculation model, and obtain the model quality characteristic value. The model quality data calculation model is specifically: ,in represents the model quality feature value, FM represents the F1 score, Te represents the total text entropy value, Sd represents the semantic vector dimension, Represents the number of words in the policy document database.
[0015] Preferably, the model training data analysis unit is used to establish a model training data calculation model, import the model training data in the policy data collection step into the model training data calculation model, and calculate the model training efficiency characteristic value. The model training data calculation model is specifically: ,in represents the model efficiency characteristic value, Ts represents the policy text processing speed, Ae represents the number of standard entities, and Rt represents the text processing time.
[0016] Preferably, the keyword data analysis unit is used to establish a keyword data calculation model, import the keyword data in the policy data collection step into the keyword data calculation model, and calculate the keyword rationality feature value. The keyword data calculation model is specifically: ,in represents the keyword rationality characteristic value, Kr represents the keyword coverage, Ks represents the keyword extraction speed, and Kn represents the number of keywords.
[0017] Preferably, the semantic similarity data analysis unit is used to establish a semantic similarity data calculation model, import the semantic similarity data in the policy data collection step into the semantic similarity data calculation model, and calculate the semantic similarity feature value. The semantic similarity data calculation model is specifically: ,in represents the semantic similarity feature value, St represents the similarity threshold, Dr represents the retrieval rate of similar policy documents, and Jt represents the retrieval response time.
[0018] Preferably, the adaptability data analysis unit is used to establish an adaptability data calculation model, import the adaptability data in the policy data collection step into the adaptability data calculation model, and calculate the adaptability characteristic value. The adaptability data calculation model is specifically: ,in represents the adaptability characteristic value, Uf represents the policy update frequency, Ur represents the policy abolition rate, and Dm represents the domain migration loss rate.
[0019] Preferably, the policy interpretation data analysis unit is used to establish a policy interpretation rationality index calculation model, and is used to import the model quality characteristic value, model training efficiency characteristic value, keyword rationality characteristic value, semantic similarity characteristic value and adaptability characteristic value of the policy data transmitted in the policy data analysis step into the policy interpretation rationality index calculation model to obtain the policy interpretation rationality index value; the policy interpretation rationality index calculation model is specifically: , where S is the policy interpretation rationality index value, represents the model quality characteristic value of policy data, Represents the model training efficiency characteristic value, Represents the keyword rationality eigenvalue, Represents the adaptability eigenvalue, Represents the semantic similarity eigenvalue.
[0020] Preferably, the preset value of the policy interpretation rationality index is marked as , S is the value of the policy interpretation rationality index. When = S, it indicates that the value of the policy interpretation rationality index is greater than the preset value of the policy interpretation rationality index, indicating that the policy interpretation is relatively reasonable; when > S, it indicates that the value of the policy interpretation rationality index is less than the preset value of the policy interpretation rationality index, indicating that the policy interpretation is unreasonable.
[0021] Preferably, a policy interpretation system based on deep learning, characterized by comprising:
[0022] Policy data collection module: used to collect model quality data, model training data, keyword data, semantic similarity data, and adaptability data of policies, and transmit the collected data to the policy data analysis module;
[0023] Policy data analysis module: includes a model quality data analysis unit, a model training data analysis unit, a keyword data analysis unit, a semantic similarity data analysis unit, and an adaptability data analysis unit. Through each analysis unit, a corresponding mathematical model is established, and the data transmitted by the policy data collection module is imported into the corresponding mathematical model to calculate the model quality eigenvalue, model training efficiency eigenvalue, keyword rationality eigenvalue, semantic similarity eigenvalue, and adaptability eigenvalue of the policy data, and transmit them to the policy interpretation data analysis module;
[0024] Policy interpretation data analysis module: includes a policy interpretation data analysis unit, used to analyze the data transmitted by the policy data analysis module, and transmit the analysis results to the final evaluation module;
[0025] Final evaluation module: used to establish a preset value of the policy interpretation rationality index, judge the policy rationality index value according to the preset value of the policy interpretation rationality index, and transmit the judgment result to the real-time feedback module;
[0026] Real-time feedback module: used to receive the judgment result transmitted by the final evaluation module. When the policy interpretation is relatively reasonable, it issues a normal signal. When the policy interpretation is unreasonable, it issues a warning signal to the terminal management personnel.
[0027] The technical effects and advantages of the present invention:
[0028] 1. The present invention collects policy-related data from multiple aspects through the policy data collection step, including policy model quality data, model training data, keyword data, semantic similarity data and adaptability data, which provides strong data support for data analysis and also reduces the adverse effects of data uniformity on analysis results;
[0029] 2. Analyze relevant policy data from multiple angles through policy data analysis steps, improve the accuracy of analysis results, and provide a theoretical basis for the rationality of policy interpretation;
[0030] 3. The analysis results are re-analyzed through the final evaluation step, which further improves the accuracy of the analysis results; problems are discovered in a timely manner through the real-time feedback step, thus providing a guarantee for the rationality of policy interpretation. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 It is a schematic diagram of the structure of the method of the present invention.
[0032] Figure 2 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION
[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0034] refer to Figure 1 A policy interpretation method based on deep learning is shown, which includes the following steps:
[0035] Step P1: Policy data collection step: used to collect policy model quality data, model training data, keyword data, semantic similarity data and adaptability data, and transfer the collected data to the policy data analysis step.
[0036] The model quality data includes the F1 score FM, the total text entropy value Te and the semantic vector dimension Sd; the model training data includes the policy text processing speed Ts, the number of standard entities Ae and the text reasoning time Rt; the keyword quality data includes the keyword coverage Kr, the keyword extraction speed Ks and the number of keywords Kn; the semantic similarity data includes the similarity threshold St, the similar policy document retrieval rate Dr and the retrieval response time Jt; the adaptability data includes the policy update frequency Uf, the policy abolition rate Ur and the domain migration loss rate Dm.
[0037] The F1 score in the model quality data is calculated by precision and recall, which is specifically expressed as: , where precision refers to the proportion of truly positive examples among those predicted by the model, and recall refers to the proportion of all truly positive examples that are correctly predicted by the model. The precision is specifically expressed as , the recall rate is specifically expressed as , where TP represents the number of samples that are truly positive and predicted as positive by the model, FP represents the number of samples that are truly negative but incorrectly predicted as positive by the model, and FN represents the number of samples that are truly positive but incorrectly predicted as negative by the model.
[0038] The specific method for collecting the total text entropy value in the model quality data is:
[0039] 1: Count the frequency of each word in the text;
[0040] 2: Calculate the probability of each word appearing;
[0041] 3: Use the entropy formula to calculate the total entropy value of the text. The entropy formula is specifically expressed as , where E represents the entropy of the text, c represents different types of classification, Represents the probability of a word.
[0042] The semantic vector dimensions in the model quality data are collected through the Word2Vec model.
[0043] The similarity threshold is collected through a recurrent neural network (RNN) model.
[0044] Step P2: Policy data analysis step: including model quality data analysis unit, model training data analysis unit, keyword data analysis unit, semantic similarity data analysis unit and adaptability data analysis unit. A corresponding mathematical model is established through each analysis unit, and the data transmitted in the policy data collection step is imported into the corresponding mathematical model. The model quality characteristic value, model training efficiency characteristic value, keyword rationality characteristic value, semantic similarity characteristic value and adaptability characteristic value of the policy data are calculated and transmitted to the policy interpretation data analysis step.
[0045] The model quality data analysis unit is used to establish a model quality data calculation model, import the model quality data in the policy data collection step into the model quality data calculation model, and obtain the model quality characteristic value. The model quality data calculation model is specifically: ,in represents the model quality feature value, FM represents the F1 score, Te represents the total text entropy value, Sd represents the semantic vector dimension, Represents the number of words in the policy document database.
[0046] The model training data analysis unit is used to establish a model training data calculation model, import the model training data in the policy data collection step into the model training data calculation model, and calculate the model training efficiency characteristic value. The model training data calculation model is specifically: ,in represents the model efficiency characteristic value, Ts represents the policy text processing speed, Ae represents the number of standard entities, and Rt represents the text processing time.
[0047] The keyword data analysis unit is used to establish a keyword data calculation model, import the keyword data in the policy data collection step into the keyword data calculation model, and calculate the keyword rationality feature value. The keyword data calculation model is specifically: ,in represents the keyword rationality characteristic value, Kr represents the keyword coverage, Ks represents the keyword extraction speed, and Kn represents the number of keywords.
[0048] The semantic similarity data analysis unit is used to establish a semantic similarity data calculation model, import the semantic similarity data in the policy data collection step into the semantic similarity data calculation model, and calculate the semantic similarity feature value. The semantic similarity data calculation model is specifically: ,in represents the semantic similarity feature value, St represents the similarity threshold, Dr represents the retrieval rate of similar policy documents, and Jt represents the retrieval response time.
[0049] The adaptability data analysis unit is used to establish an adaptability data calculation model, import the adaptability data in the policy data collection step into the adaptability data calculation model, and calculate the adaptability characteristic value. The adaptability data calculation model is specifically: ,in represents the adaptability characteristic value, Uf represents the policy update frequency, Ur represents the policy abolition rate, and Dm represents the domain migration loss rate.
[0050] Step P3: Policy interpretation data analysis step: includes a policy interpretation data analysis unit, which is used to analyze the data transmitted in the policy data analysis step and transmit the analysis results to the final evaluation step.
[0051] The policy interpretation data analysis unit is used to establish a policy interpretation rationality index calculation model, and is used to import the model quality characteristic value, model training efficiency characteristic value, keyword rationality characteristic value, semantic similarity characteristic value and adaptability characteristic value of the policy data transmitted in the policy data analysis step into the policy interpretation rationality index calculation model to obtain the policy interpretation rationality index value; the policy interpretation rationality index calculation model is specifically: , where S is the policy interpretation rationality index value, represents the model quality characteristic value of policy data, represents the model training efficiency characteristic value, represents the keyword rationality characteristic value, represents the adaptability characteristic value, represents the semantic similarity characteristic value.
[0052] Step P4: Final evaluation step: Used to establish the preset value of the policy interpretation rationality index, judge the policy rationality index value according to the preset value of the policy interpretation rationality index, and transmit the judgment result to the real-time feedback step.
[0053] The preset value of the policy interpretation rationality index is marked as , S is the policy interpretation rationality index value. When =< S, it means that the policy interpretation rationality index value is greater than the preset value of the policy interpretation rationality index, indicating that the policy interpretation is relatively reasonable; when > S, it means that the policy interpretation rationality index value is less than the preset value of the policy interpretation rationality index, indicating that the policy interpretation is unreasonable.
[0054] Step P5: Real-time feedback step: Used to receive the judgment result transmitted by the final evaluation step. When the policy interpretation is relatively reasonable, a normal signal is sent. When the policy interpretation is unreasonable, a warning signal is sent to the terminal management personnel.
[0055] Refer to Figure 2 A policy interpretation system based on deep learning shown. The present invention provides a policy interpretation system based on deep learning, including a policy data collection module, a policy data analysis module, a policy interpretation data analysis module, a final evaluation module, and a real-time feedback module.
[0056] The policy data collection module is used to collect the model quality data, model training data, keyword data, semantic similarity data, and adaptability data of the policy, and transmit the collected data to the policy data analysis module.
[0057] The policy data analysis module includes a model quality data analysis unit, a model training data analysis unit, a keyword data analysis unit, a semantic similarity data analysis unit, and an adaptability data analysis unit. By establishing corresponding mathematical models through each analysis unit, the data transmitted by the policy data collection module is imported into the corresponding mathematical models, and the model quality characteristic value, model training efficiency characteristic value, keyword rationality characteristic value, semantic similarity characteristic value, and adaptability characteristic value of the policy data are calculated and transmitted to the policy interpretation data analysis module.
[0058] The policy interpretation data analysis module includes a policy interpretation data analysis unit, which is used to analyze the data transmitted by the policy data analysis module and transmit the analysis results to the final evaluation module.
[0059] The final evaluation module is used to establish a preset value of the policy interpretation rationality index, judge the policy rationality index value according to the preset value of the policy interpretation rationality index, and transmit the judgment result to the real-time feedback module.
[0060] The real-time feedback module is used to receive the judgment result transmitted by the final evaluation module. When the policy interpretation is relatively reasonable, a normal signal is sent out; when the policy interpretation is unreasonable, an early warning signal is sent out to the terminal management personnel.
[0061] The present invention collects policy-related data from multiple aspects through the policy data collection step, including policy model quality data, model training data, keyword data, semantic similarity data and adaptability data, which provides strong data support for data analysis, and also reduces the adverse impact of data uniformity on analysis results; through the policy data analysis step, the policy-related data are analyzed from multiple angles, the accuracy of the analysis results is improved, and a theoretical basis is provided for judging the rationality of policy interpretation; through the final evaluation step, the analysis results are re-analyzed, which further improves the accuracy of the analysis results; through the real-time feedback step, problems are discovered in a timely manner, thereby providing a guarantee for the rationality of policy interpretation.
[0062] Secondly: In the drawings of the embodiments disclosed in the present invention, only the structures related to the embodiments disclosed in the present invention are involved, and other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of the present invention can be combined with each other;
[0063] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A policy interpretation method based on deep learning, characterized in that: include: Step P1: Policy data collection step: used to collect policy model quality data, model training data, keyword data, semantic similarity data and adaptability data, and transmit the collected data to the policy data analysis step; The model quality data includes F1 score FM, total text entropy value Te and semantic vector dimension Sd; model training data includes policy text processing speed Ts, standard entity quantity Ae and text reasoning time Rt; keyword quality data includes keyword coverage Kr, keyword extraction speed Ks and keyword quantity Kn; semantic similarity data includes similarity threshold St, similar policy document retrieval rate Dr and retrieval response time Jt; adaptability data includes policy update frequency Uf, policy abolition rate Ur and domain migration loss rate Dm; Step P2: Policy data analysis step: including model quality data analysis unit, model training data analysis unit, keyword data analysis unit, semantic similarity data analysis unit and adaptability data analysis unit, establishing corresponding mathematical models through each analysis unit, importing the data transmitted in the policy data collection step into the corresponding mathematical model, calculating the model quality characteristic value, model training efficiency characteristic value, keyword rationality characteristic value, semantic similarity characteristic value and adaptability characteristic value of the policy data, and transmitting them to the policy interpretation data analysis step; The model quality data analysis unit is used to establish a model quality data calculation model, import the model quality data in the policy data collection step into the model quality data calculation model, and obtain the model quality characteristic value. The model quality data calculation model is specifically: ,in represents the model quality feature value, FM represents the F1 score, Te represents the total text entropy value, Sd represents the semantic vector dimension, represents the number of words in the policy document database; The model training data analysis unit is used to establish a model training data calculation model, import the model training data in the policy data collection step into the model training data calculation model, and calculate the model training efficiency characteristic value. The model training data calculation model is specifically: ,in represents the model efficiency characteristic value, Ts represents the policy text processing speed, Ae represents the number of standard entities, and Rt represents the text processing time; The keyword data analysis unit is used to establish a keyword data calculation model, import the keyword data in the policy data collection step into the keyword data calculation model, and calculate the keyword rationality feature value. The keyword data calculation model is specifically: ,in represents the keyword rationality characteristic value, Kr represents the keyword coverage rate, Ks represents the keyword extraction speed, and Kn represents the number of keywords; The semantic similarity data analysis unit is used to establish a semantic similarity data calculation model, import the semantic similarity data in the policy data collection step into the semantic similarity data calculation model, and calculate the semantic similarity feature value. The semantic similarity data calculation model is specifically: ,in represents the semantic similarity feature value, St represents the similarity threshold, Dr represents the retrieval rate of similar policy documents, and Jt represents the retrieval response time; The adaptability data analysis unit is used to establish an adaptability data calculation model, import the adaptability data in the policy data collection step into the adaptability data calculation model, and calculate the adaptability characteristic value. The adaptability data calculation model is specifically: ,in represents the adaptive characteristic value, Uf represents the policy update frequency, Ur represents the policy abolition rate, and Dm represents the domain migration loss rate; Step P3: Policy interpretation data analysis step: including a policy interpretation data analysis unit, which is used to analyze the data transmitted in the policy data analysis step and transmit the analysis results to the final evaluation step; Step P4: Final evaluation step: used to establish a preset value of the policy interpretation rationality index, judge the policy rationality index value according to the preset value of the policy interpretation rationality index, and transmit the judgment result to the real-time feedback step; Step P5: Real-time feedback step: used to receive the judgment result transmitted by the final evaluation step. When the policy interpretation is reasonable, a normal signal is sent; when the policy interpretation is unreasonable, an early warning signal is sent to the terminal management personnel.
2. According to claim 1, a policy interpretation method based on deep learning is characterized by: The policy interpretation data analysis unit is used to establish a policy interpretation rationality index calculation model, and is used to import the model quality characteristic value, model training efficiency characteristic value, keyword rationality characteristic value, semantic similarity characteristic value and adaptability characteristic value of the policy data transmitted in the policy data analysis step into the policy interpretation rationality index calculation model to obtain the policy interpretation rationality index value; the policy interpretation rationality index calculation model is specifically: , where S is the policy interpretation rationality index value, represents the model quality characteristic value of policy data, Represents the model training efficiency characteristic value, Indicates the keyword rationality feature value, represents the adaptive characteristic value, Represents the semantic similarity feature value.
3. According to claim 1, a policy interpretation method based on deep learning is characterized by: The preset value of the policy interpretation rationality index is marked as , where S is the value of the policy interpretation rationality index. When =<S, it indicates that the value of the policy interpretation rationality index is greater than the preset value of the policy interpretation rationality index, indicating that the policy interpretation is relatively reasonable; when >S, it indicates that the value of the policy interpretation rationality index is less than the preset value of the policy interpretation rationality index, indicating that the policy interpretation is unreasonable.
4. A policy interpretation system based on deep learning, characterized in that: include: Policy data collection module: used to collect policy model quality data, model training data, keyword data, semantic similarity data and adaptability data, and transmit the collected data to the policy data analysis module; The model quality data includes F1 score FM, total text entropy value Te and semantic vector dimension Sd; model training data includes policy text processing speed Ts, standard entity quantity Ae and text reasoning time Rt; keyword quality data includes keyword coverage Kr, keyword extraction speed Ks and keyword quantity Kn; semantic similarity data includes similarity threshold St, similar policy document retrieval rate Dr and retrieval response time Jt; adaptability data includes policy update frequency Uf, policy abolition rate Ur and domain migration loss rate Dm; Policy data analysis module: including model quality data analysis unit, model training data analysis unit, keyword data analysis unit, semantic similarity data analysis unit and adaptability data analysis unit. Corresponding mathematical models are established through each analysis unit, and the data transmitted by the policy data acquisition module is imported into the corresponding mathematical model to calculate the model quality characteristic value, model training efficiency characteristic value, keyword rationality characteristic value, semantic similarity characteristic value and adaptability characteristic value of the policy data, and transmit them to the policy interpretation data analysis module; The model quality data analysis unit is used to establish a model quality data calculation model, import the model quality data in the policy data collection step into the model quality data calculation model, and obtain the model quality characteristic value. The model quality data calculation model is specifically: ,in represents the model quality feature value, FM represents the F1 score, Te represents the total text entropy value, Sd represents the semantic vector dimension, represents the number of words in the policy document database; The model training data analysis unit is used to establish a model training data calculation model, import the model training data in the policy data collection step into the model training data calculation model, and calculate the model training efficiency characteristic value. The model training data calculation model is specifically: ,in represents the model efficiency characteristic value, Ts represents the policy text processing speed, Ae represents the number of standard entities, and Rt represents the text processing time; The keyword data analysis unit is used to establish a keyword data calculation model, import the keyword data in the policy data collection step into the keyword data calculation model, and calculate the keyword rationality feature value. The keyword data calculation model is specifically: ,in represents the keyword rationality characteristic value, Kr represents the keyword coverage rate, Ks represents the keyword extraction speed, and Kn represents the number of keywords; The semantic similarity data analysis unit is used to establish a semantic similarity data calculation model, import the semantic similarity data in the policy data collection step into the semantic similarity data calculation model, and calculate the semantic similarity feature value. The semantic similarity data calculation model is specifically: ,in represents the semantic similarity feature value, St represents the similarity threshold, Dr represents the retrieval rate of similar policy documents, and Jt represents the retrieval response time; The adaptability data analysis unit is used to establish an adaptability data calculation model, import the adaptability data in the policy data collection step into the adaptability data calculation model, and calculate the adaptability characteristic value. The adaptability data calculation model is specifically: ,in represents the adaptive characteristic value, Uf represents the policy update frequency, Ur represents the policy abolition rate, and Dm represents the domain migration loss rate; Policy interpretation data analysis module: including a policy interpretation data analysis unit, which is used to analyze the data transmitted by the policy data analysis module and transmit the analysis results to the final evaluation module; The policy interpretation data analysis unit is used to establish a policy interpretation rationality index calculation model, and is used to import the model quality characteristic value, model training efficiency characteristic value, keyword rationality characteristic value, semantic similarity characteristic value and adaptability characteristic value of the policy data transmitted in the policy data analysis step into the policy interpretation rationality index calculation model to obtain the policy interpretation rationality index value; the policy interpretation rationality index calculation model is specifically: , where S is the policy interpretation rationality index value, represents the model quality characteristic value of policy data, Represents the model training efficiency characteristic value, Indicates the keyword rationality feature value, represents the adaptive characteristic value, Represents the semantic similarity feature value; Final evaluation module: used to establish a preset value of the policy interpretation rationality index, judge the policy rationality index value according to the preset value of the policy interpretation rationality index, and transmit the judgment result to the real-time feedback module; Real-time feedback module: used to receive the judgment results transmitted by the final evaluation module. When the policy interpretation is relatively reasonable, it sends a normal signal. When the policy interpretation is unreasonable, it sends an early warning signal to the terminal management personnel.
Citation Information
Patent Citations
Policy interpretation method based on deep learning
CN112541352A
Military and civilian fusion policy information data analysis and optimization system
CN113254512A
User feedback data analysis method and device and computer equipment
CN115114916A
Public data platform based on multi-party data sources and data analysis method thereof
CN117574915A
KR20230136396A
