Administrative law enforcement auxiliary system based on large language model and retrieval enhancement generation technology

Abstract

The invention provides an administrative law enforcement auxiliary system based on a large language model and a retrieval enhancement generation technology, and relates to the technical field of artificial intelligence, comprising: an analysis processing module obtains case information of a law enforcement scene, and understands and analyzes the case information through the large language model to obtain case description information; a retrieval analysis module performs associated information retrieval in a knowledge base according to the case description information based on a retrieval enhancement generation technology to obtain target analysis data; and the suggestion feedback module determines law enforcement suggestions according to the target analysis data and performs feedback display on the law enforcement suggestions in combination with the target analysis data. According to the method, a large language model and a retrieval enhancement generation technology are combined, so that law enforcement officers can efficiently look up related data by means of artificial intelligence in the actual law enforcement process, meanwhile, dependence on the ability of the law enforcement officers can be reduced, the accuracy and interpretability of law enforcement bases are improved, the decision accuracy is ensured, and the law enforcement efficiency is improved. And the fairness and consistency of law enforcement are ensured.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an administrative law enforcement assistance system based on a large language model and retrieval enhancement generation technology. Background Technology

[0002] When handling various cases, law enforcement officers need to rely on a large number of laws, regulations, policy documents, and historical cases to make accurate judgments and decisions. However, in current law enforcement work, officers often need to manually consult a large number of laws, regulations, policy documents, and historical cases. This process is cumbersome, consumes a lot of time searching through knowledge bases, is inefficient, wastes a lot of manpower and time, may lead to delays in law enforcement, and is also prone to overlooking key evidence, affecting the accuracy and timeliness of law enforcement. Moreover, different law enforcement officers may have different understandings of the same legal provisions due to factors such as professional background and experience level, resulting in inconsistent enforcement standards. Furthermore, when handling cases, law enforcement officers need to deeply understand and analyze the case content according to their own level of understanding. However, the varying abilities of law enforcement officers can easily lead to biased law enforcement decisions, resulting in misjudgments or inconsistent enforcement in actual law enforcement, affecting the fairness and consistency of law enforcement, and may also damage the authority of law enforcement, causing public questioning of the impartiality of law enforcement.

[0003] Therefore, a new administrative law enforcement support system is urgently needed to solve the above problems. Summary of the Invention

[0004] The purpose of this invention is to provide an administrative law enforcement assistance system based on large language models and retrieval enhancement generation technology. By combining large language models and retrieval enhancement generation technology, law enforcement officers can use artificial intelligence to analyze and retrieve laws and regulations during actual law enforcement. This not only saves manpower and time and efficiently realizes the retrieval of relevant information, but also reduces reliance on the capabilities of law enforcement officers, improves the accuracy and interpretability of the law enforcement basis, ensures the accuracy of decision-making, and guarantees the fairness and consistency of law enforcement.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an administrative law enforcement auxiliary system based on a large language model and retrieval enhancement generation technology, comprising: The analysis and processing module is used to acquire case information from law enforcement scenarios and to understand and analyze the case information through a large language model to obtain case description information. The retrieval and analysis module is used to retrieve related information from the knowledge base based on the case description information and retrieval enhancement generation technology to obtain target analysis data. The suggestion feedback module is used to determine law enforcement recommendations based on target analysis data, and then display the law enforcement recommendations in conjunction with the target analysis data.

[0006] Preferably, the analysis and processing module includes: The acquisition unit is used to acquire case information for law enforcement scenarios; The processing unit is used to understand and analyze case information using a large language model, including: preprocessing the case information to obtain first-processed information; and using a standardized processing model to vectorize the first-processed information in combination with the context to obtain case description information.

[0007] Preferably, the acquisition unit also acquires feature information of law enforcement personnel to obtain law enforcement personnel feature information, thereby optimizing the case information by combining the law enforcement personnel feature information before the processing unit performs information preprocessing on the case information.

[0008] Preferably, before the processing unit performs understanding and analysis of case information using the large language model, the large language model is fine-tuned according to the field of administrative law enforcement.

[0009] Preferably, the retrieval and analysis module includes: The retrieval and analysis unit is used to configure the indexing mechanism for the knowledge base and perform associated indexing in the corresponding knowledge base according to the case description information based on the matching indexing mechanism to obtain the target associated information; The results generation unit is used to enhance the generation of search results based on target association information to obtain target analysis data.

[0010] Preferably, the associated information includes: first associated information, second associated information, and third associated information; When the result generation unit enhances the generation of search results based on target association information, it organizes the first association information in conjunction with the third association information to obtain the first target analysis data, and organizes it based on the second association information to obtain the second target analysis data.

[0011] Preferably, the retrieval and analysis module includes: an update management unit; The update management unit is used to dynamically update the knowledge base, and at the same time, to manage the version of the knowledge base as the version is updated.

[0012] Preferably, the knowledge base includes: a legal and regulatory database, a historical case database, and a policy document database. When the update management unit performs dynamic knowledge updates on the knowledge base, it also performs regular updates on the historical case database. The legal and regulatory database and the policy document database are monitored for adjustments, and the updated version is determined based on the adjustment monitoring information. Then, the updated version is scheduled for a specific time based on the implementation time.

[0013] Preferably, the suggestion feedback module includes: The suggestion generation unit is used to determine law enforcement recommendations based on the analysis of the primary objective data. The feedback display unit is used to display the law enforcement recommendations in conjunction with the target analysis data. It displays the main feedback for the law enforcement recommendations and additional feedback for the target analysis data through hidden links.

[0014] Preferably, the suggestion feedback module further includes: an interactive query unit; The interactive query unit is used to respond to law enforcement officers' requests for further information regarding the feedback displayed information.

[0015] The present invention has achieved the following beneficial effects: This invention utilizes a large language model and retrieval-enhanced generation technology to provide more convenient services for law enforcement personnel in administrative law enforcement support systems. It enables personnel to better understand case information during actual enforcement processes through the large language model, and accurately retrieve related information from the knowledge base based on retrieval-enhanced generation technology. This not only reduces the time and effort required by law enforcement personnel, efficiently and comprehensively accessing relevant materials and avoiding omissions of key evidence, but also quickly identifies more accurate target analysis data, improving the accuracy and interpretability of enforcement evidence. Furthermore, it reduces reliance on the capabilities of law enforcement personnel, allowing them to determine target analysis data and enforcement recommendations without requiring specialized identification and reasoning. This avoids biases in enforcement decisions due to varying capabilities among personnel, ensuring fairness and consistency in enforcement, and safeguarding the authority of law enforcement. Simultaneously, during feedback display, not only enforcement recommendations are shown, but target analysis data is also displayed synchronously, enhancing the transparency and traceability of enforcement decisions.

[0016] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the application.

[0017] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the administrative law enforcement auxiliary system described in this invention; Figure 2 This is a schematic diagram of the analysis and processing module in the administrative law enforcement auxiliary system described in this invention; Figure 3 This is a schematic diagram of the retrieval and analysis module in the administrative law enforcement auxiliary system described in this invention; Figure 4 This is a schematic diagram of the suggestion feedback module in the administrative law enforcement auxiliary system described in this invention. Detailed Implementation

[0019] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0020] like Figure 1 As shown, this embodiment of the invention provides an administrative law enforcement assistance system based on a large language model and retrieval enhancement generation technology, including: an analysis and processing module, a retrieval analysis module, and a suggestion feedback module. The analysis and processing module, the retrieval analysis module, and the suggestion feedback module are connected sequentially.

[0021] The analysis and processing module is used to acquire case information from law enforcement scenarios and to understand and analyze the case information using a large language model to obtain case description information.

[0022] Among them, the large language model is an artificial intelligence model built on deep learning technology. Through the Transformer architecture, it combines case information imported by law enforcement officers with language patterns, structures, and semantics for understanding and analysis, thereby generating natural language text or understanding and processing natural language text. Here, the large language model is pre-trained using large-scale text data for cases in various law enforcement scenarios.

[0023] The retrieval and analysis module is used to retrieve related information from the knowledge base based on case description information and retrieval enhancement generation technology to obtain target analysis data.

[0024] The knowledge base is a database that stores external knowledge, including: a legal and regulatory database, a historical case database, and a policy document database. Related information refers to the laws and regulations involved in the case, similar historical cases, and relevant policy documents.

[0025] The retrieval enhancement generation technology is a retrieval analysis method that deeply integrates a large language model with a knowledge base through the RAG architecture. Based on the output information of the large language model (i.e., case description information), it performs retrieval queries in the knowledge base (legal and regulatory database, historical case database, policy document database, etc.) and then uses the information retrieved to enhance and generate retrieval results.

[0026] The suggestion feedback module is used to determine law enforcement recommendations based on target analysis data, and then display the law enforcement recommendations in conjunction with the target analysis data.

[0027] When combining enforcement recommendations with target analysis data for feedback display, not only are the enforcement recommendations displayed, but also related information such as the legal provisions on which they are based, applicable conditions, and similar cases are also displayed simultaneously.

[0028] The aforementioned technical solution, through large language models and retrieval-enhanced generation technology, enables administrative law enforcement auxiliary systems to provide more convenient services for law enforcement personnel. It allows personnel to better understand case information during actual enforcement, leveraging large language models to accurately retrieve related information from the knowledge base based on retrieval-enhanced generation technology. This not only reduces the time and effort required by law enforcement personnel, efficiently and comprehensively accessing relevant materials and avoiding omissions of key evidence, but also quickly identifies more accurate target analysis data, improving the accuracy and interpretability of enforcement evidence. Furthermore, it reduces reliance on the capabilities of law enforcement personnel, eliminating the need for specialized identification and reasoning to determine target analysis data and enforcement recommendations. This avoids biases in enforcement decisions due to varying skill levels among personnel, ensuring fairness and consistency in enforcement and safeguarding its authority. Simultaneously, during feedback display, not only enforcement recommendations are shown, but target analysis data is also displayed synchronously, enhancing the transparency and traceability of enforcement decisions.

[0029] In one embodiment provided by the present invention, such as Figure 2 As shown, the analysis and processing module includes an acquisition unit and a processing unit.

[0030] The acquisition unit is used to acquire case information for law enforcement scenarios.

[0031] The case information includes the location of the incident and information about the crime scene. The acquisition unit employs diverse methods to obtain case information in law enforcement scenarios, including text input, voice recording, and image recognition.

[0032] The processing unit is used to understand and analyze case information using a large language model, including: preprocessing the case information to obtain first-processed information; and using a standardized processing model to vectorize the first-processed information in combination with the context to obtain case description information.

[0033] Among them, information preprocessing includes: text cleaning, word segmentation,词性标注, key entity recognition, etc. When performing information preprocessing on case information, text cleaning is carried out on the case information to remove irrelevant characters, punctuation marks, stop words, etc. in the text expression of the case information, so as to ensure the cleanliness and standardization of the text of the case information and obtain the first preprocessed information. Then, analyze the sentence structure of the first preprocessed information and perform word segmentation according to the sentence structure to split the continuous text expression into independent words, obtaining the second preprocessed information. For example: The street vendor sets up a stall on the sidewalk at the intersection of XX Road and XX Avenue. After word segmentation, we get: street vendor / set up a stall / on / the sidewalk / at / the intersection of / XX Road / and / XX Avenue. Next, perform词性分析与标注 on the second preprocessed information. For example: The word "street vendor" is a noun, "set up a stall" is a verb, "on" is a preposition, "the sidewalk" is a noun, "at" is a preposition, "the intersection of" is a noun phrase, "XX Road" is a proper noun, "and" is a conjunction, "XX Avenue" is a proper noun. At the same time, identify and label the key entities in the second preprocessed information, such as: personal names, place names, organization names, time, etc. Further obtain the first processed information about the case information.

[0034] Here, the normalization processing model is a vectorization model adopting an encoder-decoder architecture, which converts text information into a vector form for output, such as: Transformer. When using the normalization processing model to perform vectorization processing on the first processed information in combination with the context, it includes: mapping the words in the first processed information into a low-dimensional vector space respectively to obtain word vectors; determining the self-attention mechanism according to the annotation information in the first processed information; performing word interaction on the word vectors in combination with the self-attention mechanism to obtain global context information and get context vectors; initializing the decoder according to the context vectors, and the decoder with the initialized state gradually generates a sequence vector of case information descriptions according to the context vectors to obtain case description information.

[0035] It should be noted that the "词性标注" and "词性分析与标注" in the original text need to be replaced with the accurate English expressions according to the specific content. Since the specific English terms are not provided in the original text, the above translation uses the original Chinese terms for now.The aforementioned technical solution, through its processing unit, enables the analysis and processing module to promptly understand and analyze case information after the acquisition unit obtains it. This avoids discrepancies in the textual expression of case information that could prevent the retrieval and analysis module from accurately retrieving related information, thus ensuring the reliability of the retrieval and analysis model. Furthermore, the processing unit, through information preprocessing, standardizes the case information, ensuring it has a unified data format. This guarantees the retrieval and analysis module's ability to retrieve related information, preventing inconsistencies that could prevent effective retrieval of certain case information. This improves the comprehensiveness and accuracy of the target analysis data. Moreover, by employing a standardized processing model and combining the first-processed information with contextual vectorization, the case description information is presented as a vector in the retrieval and analysis module. Based on retrieval enhancement generation technology, related information is retrieved from the knowledge base, ensuring the retrieval of related information and improving its efficiency. This allows for the rapid identification of target analysis data, reducing feedback display waiting time and enabling law enforcement personnel to make efficient law enforcement decisions, thereby improving the efficiency of law enforcement processing.

[0036] In one embodiment of the present invention, the acquisition unit further acquires feature information of law enforcement personnel to obtain law enforcement personnel feature information, thereby optimizing the case information by combining the law enforcement personnel feature information before the processing unit performs information preprocessing on the case information.

[0037] Among them, law enforcement personnel characteristic information refers to the expressive and behavioral characteristics of law enforcement personnel, such as dialectal or simplified expressions. Before the processing unit performs information preprocessing on case information, information optimization based on law enforcement personnel characteristic information includes: A preliminary analysis of case information is conducted using a statement analysis model to determine whether the textual expression of the case information is standardized. In this step, the expression analysis model is a text expression recognition and analysis model based on standardized text expression. During the preliminary analysis of case information, the expression analysis model is used to analyze whether the text expression of the case information is expressed using standardized language habits, thereby determining whether the text expression of the case information is standardized.

[0038] When the textual expression of case information is standardized, the system analyzes the characteristics of law enforcement officers to determine whether they have a habit of using dialects. If they do, the system obtains their dialect characteristics and then further analyzes the case information based on these characteristics to determine whether dialect expressions exist in the textual expression. If so, the system optimizes the expression to obtain an optimized textual expression of the case information.

[0039] In this step, if law enforcement officers do not have dialectal expression habits, there is no need to optimize the case information. Dialectal expression includes dialectal vocabulary and dialectal expression habits, such as reverse order. When optimizing dialectal expression, for dialectal vocabulary, the vocabulary is converted into standardized expressions, and these standardized expressions are used to replace the dialectal vocabulary in the case information to obtain an optimized case information text; for dialectal expression habits, the order of expression in the case information is adjusted to obtain an optimized case information text. Here, standardized expression refers to formal expression that the public can understand.

[0040] When the textual expression of case information is not standardized, the simplified expression habits of law enforcement officers are obtained based on their characteristic information. The case information is then subjected to a first optimization process based on these simplified expression habits, which refines and optimizes the simplified expression in the case information to obtain the first optimized information. At the same time, the law enforcement officers are analyzed to determine whether they have dialectal expression habits. If they do, their dialectal characteristics are obtained. Then, the first optimized information is further analyzed based on these dialectal characteristics to determine whether dialectal expressions exist in the textual expression of the first optimized information. If so, a second optimization process is performed on the first optimized information, replacing the dialectal expressions in the first optimized information with standardized expressions to obtain the optimized textual expression of the case information.

[0041] In this step, if law enforcement officers do not have a habit of using dialects, the first optimized information is the optimized expression of the case information text.

[0042] The aforementioned acquisition of law enforcement personnel characteristic information ensures accurate responses from different law enforcement personnel when they input case information into the administrative law enforcement assistance system. This avoids discrepancies in case information due to differences in the expression and behavioral habits of law enforcement personnel, thus ensuring the accuracy of the displayed feedback information and guaranteeing the accuracy of case descriptions. Furthermore, when optimizing case information by incorporating law enforcement personnel characteristic information, a pre-analysis model is used to analyze the case information, allowing for different processing based on the specific circumstances of the case. This improves the effectiveness of information optimization and adapts to different law enforcement personnel characteristics, thereby expanding the applicability of the administrative law enforcement assistance system and enabling accurate responses from law enforcement personnel in any region.

[0043] In one embodiment of the present invention, before the processing unit performs understanding and analysis of case information using a large language model, the large language model is fine-tuned according to the field of administrative law enforcement.

[0044] Among these adjustments, the large language model was fine-tuned according to the field of administrative law enforcement, including: Based on case information, analyze the areas of administrative law enforcement to identify target areas of administrative law enforcement.

[0045] In this step, when analyzing the administrative law enforcement field based on case information, the case type is determined based on the case information, and then the administrative law enforcement field to which the case belongs is determined based on the case type, thereby clarifying the target administrative law enforcement field.

[0046] Training information is acquired according to the target administrative law enforcement areas.

[0047] In this step, within the target administrative law enforcement field, corpus information is acquired for different cases in different law enforcement scenarios to obtain training information.

[0048] Fine-tuning of the large language model based on training information yields the fine-tuned large language model.

[0049] In this step, the large language model is trained and optimized based on the training information, the parameters of the large language model are adjusted, the semantic understanding ability of the large language model in the target administrative law enforcement field is improved, and the fine-tuned large language model is determined.

[0050] The aforementioned fine-tuning of the large language model according to administrative law enforcement domains enables the adjusted model to more accurately and clearly express information when understanding and analyzing case information, thus improving the accuracy of case descriptions. Furthermore, by identifying the target administrative law enforcement domain, the fine-tuning of the large language model can be targeted, enhancing its ability to understand case information within that domain. This leads to more accurate case descriptions, providing a basis for related information retrieval and improving the precision of target analysis data and enforcement recommendations.

[0051] In one embodiment provided by the present invention, such as Figure 3 As shown, the retrieval and analysis module includes a retrieval and analysis unit and a result generation unit.

[0052] The retrieval and analysis unit is used to configure the indexing mechanism for the knowledge base and, according to the matching indexing mechanism, index the relevant information in the corresponding knowledge base based on the case description information to obtain the target related information.

[0053] The knowledge base includes: a legal and regulatory database, a historical case database, and a policy document database. The target related information includes: primary related information, secondary related information, and tertiary related information. Different knowledge bases correspond to different indexing mechanisms. When indexing based on case description information in the corresponding knowledge base according to the matching indexing mechanism, the primary indexing mechanism is used to index the primary related information in the legal and regulatory database based on the case description information, yielding the primary related information; the secondary indexing mechanism is used to index the secondary related information in the historical case database based on the case description information, yielding the secondary related information; and the tertiary indexing mechanism is used to index the tertiary related information in the policy document database based on the case description information, yielding the tertiary related information. Here, the primary, secondary, and tertiary related information indexing can be performed simultaneously.

[0054] Furthermore, when indexing the first related information in the legal and regulatory database based on the case description information according to the first indexing mechanism, the laws and regulations in the legal and regulatory database are converted into vector representations to obtain legal and regulatory vectors; a similarity comparison analysis is performed based on the case description information and the legal and regulatory vectors to obtain similarity analysis data; the similarity analysis data is sorted, and the similarity analysis data is truncated according to a first threshold, and then the laws and regulations corresponding to the truncated similarity analysis data are obtained to obtain the first related information. When indexing the second related information in the historical case database based on the case description information according to the second indexing mechanism, key features of the case are obtained based on the case description information; feature matching is performed in the historical case database based on the key features of the case to obtain the key feature matching results; the overall relevance of historical cases is determined based on the case feature matching results; the overall relevance of historical cases is sorted, and the overall relevance is truncated according to a second threshold, and then the historical cases corresponding to the truncated overall relevance are obtained to obtain the second related information. When performing third-level indexing of policy document database based on case description information according to the third-level indexing mechanism, the case location and on-site information are determined based on the case description information. A preliminary screening is then performed in the policy document database based on the case location to obtain the first screening result. Next, policy matching analysis is performed on the first screening result based on the on-site information. Finally, target policy documents are selected based on the policy matching analysis data combined with a third threshold to obtain the third-level related information. Here, the first, second, and third thresholds may be the same or different.

[0055] The results generation unit is used to enhance the generation of search results based on target association information to obtain target analysis data.

[0056] In the process of enhancing the generation of search results based on target-related information, the target-related information is combined with case information to create an enhanced prompt. At the same time, information is organized based on the first, second, and third related information to determine the laws and regulations that law enforcement officers rely on when enforcing the law on the case corresponding to the case information. Based on the applicable laws and regulations, the applicable conditions of the laws and regulations, as well as the enforcement guidance and operational details, are obtained. Historical cases similar to the case corresponding to the case information are also identified to obtain target analysis data.

[0057] The aforementioned retrieval and analysis unit enables the retrieval of related information across different knowledge bases using different indexing mechanisms. This not only improves the efficiency of related retrieval, allowing for quick retrieval of target related information, but also enhances the flexibility and performance of related retrieval. It allows for on-demand related retrieval across different knowledge bases, ensuring the effectiveness and reliability of target analysis data. This avoids the inclusion of excessive interfering information in the target analysis data due to a single, fixed retrieval mechanism, which could lead to biased enforcement recommendations. Furthermore, the knowledge base integrates legal and regulatory databases, historical case databases, and policy document databases to fully analyze cases corresponding to the case information. This eliminates the need for law enforcement personnel to manually analyze the data in conjunction with laws and regulations, reducing the professional skill requirements for law enforcement personnel and avoiding analytical biases caused by varying levels of competence among them. This ensures the accuracy of the target analysis data and enriches its content. It not only provides law enforcement personnel with the applicable laws and regulations but also offers similar historical cases and the applicable conditions of the applicable laws and regulations, thereby ensuring the accuracy and interpretability of the enforcement basis. Moreover, by creating enhanced prompts within the target analysis data, the retrieval of case information becomes a coherent process, improving the traceability of the interpretation and analysis process of case information in the administrative law enforcement auxiliary system.

[0058] In one embodiment of the present invention, the associated information includes: first associated information, second associated information, and third associated information; When the result generation unit enhances the generation of search results based on target association information, it organizes the first association information in conjunction with the third association information to obtain the first target analysis data, and organizes it based on the second association information to obtain the second target analysis data.

[0059] The process of organizing the first and third related information includes: determining the applicable laws and regulations based on the first related information, obtaining the target laws and regulations, and determining the number of target laws and regulations. When there is only one target law and regulation, the policy documents in the third related information are combined with the target law and regulation to obtain the first target analysis data. When there are two or more target laws and regulations, the third related information is divided into related categories according to the target laws and regulations, the policy documents are bound to the target laws and regulations, and the target laws and regulations are statistically summarized one by one to obtain the first target analysis data.

[0060] The data is organized based on the second related information, including: determining the number of target historical cases based on the second related information; when there are two or more target historical cases, comparing and analyzing the target historical cases to determine the similarities and differences between them; filtering the target historical cases based on the similarities and differences to determine representative historical cases; and then summarizing and organizing the representative historical cases to obtain the second target analysis data.

[0061] The above-mentioned analysis and organization of the first and third related information provides convenience for the suggestion feedback module to determine law enforcement recommendations. This allows for direct integration of relevant policy documents and target laws and regulations when determining law enforcement recommendations, ensuring the accuracy of law enforcement and avoiding disputes arising from conflicts between policy documents and target laws and regulations. Furthermore, by organizing relevant historical cases based on the second related information, relevant historical cases are clarified, thereby providing reference for law enforcement personnel, improving the reliability of law enforcement, and ensuring the authority of law enforcement.

[0062] In one embodiment of the present invention, the retrieval and analysis module includes: an update management unit; The update management unit is used to dynamically update the knowledge base, and at the same time, to manage the version of the knowledge base as the version is updated.

[0063] Specifically, when the update management unit performs dynamic knowledge updates to the knowledge base, it performs dynamic knowledge updates to the legal and regulatory database, historical case database, and policy document database separately. Furthermore, when managing the version of the knowledge base, it records the version updates and archives the old version when the new version is implemented, thus ensuring the traceability of the old version and enabling retrieval and verification of the old version when necessary.

[0064] The aforementioned technical solution updates the knowledge base through an update management unit, ensuring timely updates when laws, regulations, or policies are adjusted. This prevents discrepancies between the knowledge base and these changes from impacting case analysis, improves the accuracy of the administrative law enforcement support system, avoids enforcement risks due to untimely knowledge base updates, and safeguards the authority of law enforcement. Furthermore, version management ensures the traceability of legal and policy adjustments, preventing misunderstandings arising from enforcement discrepancies caused by these changes.

[0065] In one embodiment of the present invention, the knowledge base includes: a legal and regulatory database, a historical case database, and a policy document database. When the update management unit performs dynamic knowledge updates on the knowledge base, it also performs regular updates on the historical case database. The legal and regulatory database and the policy document database are monitored for adjustments, and the updated version is determined based on the adjustment monitoring information. Then, the updated version is set to be updated on a timed basis in conjunction with the implementation time.

[0066] Specifically, when regularly updating the historical case database, the update frequency is set in advance, and then the database is updated regularly according to the set update frequency. When the legal and regulatory database or policy document database is updated, an emergency update command is generated for the historical case database. The historical case database is updated based on the emergency update command, and the regular update according to the set update frequency is restarted.

[0067] The system monitors and adjusts the legal and regulatory database and policy document database. Based on the monitoring information, it determines the updated version and schedules the updated version according to the implementation time. It also monitors and adjusts the database in real time for the promulgation of new laws, regulations, and / or policies. When the monitoring information indicates the promulgation of new laws, regulations, and / or policies, it determines the update information for the legal and regulatory database and / or policy document database based on the new laws, regulations, and / or policies. A new version of the legal and regulatory database and / or policy document database is then obtained based on this information. Simultaneously, the implementation time of the new laws, regulations, and / or policies is acquired. A schedule is then set for the new version of the legal and regulatory database and / or policy document database based on the implementation time. When the scheduled time condition is met, the new version of the legal and regulatory database and / or policy document database replaces the current one, and the new version is implemented, thus completing the dynamic knowledge update of the legal and regulatory database and / or policy document database.

[0068] The aforementioned technical solution regularly updates the historical case database, enriching the database and preventing the credibility of outdated cases from being compromised. This ensures the timeliness and reliability of historical cases. Furthermore, when the legal or policy document database is updated, an emergency update command promptly updates the historical cases, ensuring the system can respond instantly to rapidly changing legal or policy changes. This avoids knowledge lag caused by waiting for the next regular update cycle, guaranteeing the timeliness of the historical case database. Simultaneously, restarting regular updates according to the set update frequency avoids unnecessary follow-up updates after an emergency update, optimizing system resources.

[0069] When updating the legal and regulatory database and policy document database, the updated versions of the legal and regulatory database and policy document database are determined immediately based on the adjusted monitoring information, but do not take effect immediately. This avoids system chaos during the gap between the "promulgation date" and the "implementation date" of laws, regulations or policies, and allows for a zero-delay switch at the implementation time. This ensures the continuity and compliance of business operations, achieves a seamless compliance transition, and ensures the timeliness and authority of the enforcement basis, avoiding enforcement risks caused by changes in regulations.

[0070] In one embodiment provided by the present invention, such as Figure 4 As shown, the suggestion feedback module includes a suggestion generation unit and a feedback display unit.

[0071] The suggestion generation unit is used to determine law enforcement recommendations based on the analysis of the primary objective data.

[0072] The recommendation generation unit, when determining enforcement recommendations based on the first target analysis data, analyzes the hierarchical types of the relevant policy documents according to the target laws and regulations. These hierarchical types include national policies and local policies. When the policy document is a national policy, it interprets the target laws and regulations to determine applicable conditions, enforcement guidance, and operational details, and generates enforcement recommendations based on these conditions. When the policy document is a local policy, it analyzes whether the content of the policy document conflicts with the provisions of the target laws and regulations. If there is a conflict, it determines the applicable conditions, enforcement guidance, and operational details based on the target laws and regulations, and generates a first enforcement recommendation. This first enforcement recommendation is then adjusted based on the policy document to obtain a second enforcement recommendation, which is the final enforcement recommendation. If there is no conflict, it interprets the target laws and regulations based on the policy document to determine applicable conditions, enforcement guidance, and operational details, and generates enforcement recommendations based on these conditions.

[0073] The feedback display unit is used to display the law enforcement recommendations in conjunction with the target analysis data. It displays the main feedback for the law enforcement recommendations and additional feedback for the target analysis data through hidden links.

[0074] Specifically, when the feedback display unit displays the main feedback for law enforcement recommendations, it directly presents the recommendations as text information. When displaying supplementary feedback for target analysis data using hidden links, it identifies link-related keywords for both the first and second target analysis data, marks these keywords in the main feedback display, and then generates hidden links based on these keywords for both the first and second target analysis data. These links include: Based on the analysis data of the first target and the analysis data of the second target, a hidden information page is generated respectively, resulting in the first hidden page and the second hidden page; Match the first and second hidden pages with related keywords. Configure the domain name based on the matching results, and generate entry and landing links according to the configured domain name; Generate script files for entry and landing links; Generate patch links according to the script file, so that the associated link words can link to the corresponding first hidden page or second hidden page.

[0075] The aforementioned technical method determines enforcement recommendations based on the analysis of the first target data. This ensures that the generated recommendations incorporate both the target laws and regulations and take into account regional differences in laws and regulations caused by policy documents, guaranteeing the rationality of the recommendations in the area where the case occurred and improving the flexibility of the administrative enforcement support system. Furthermore, the feedback display unit integrates the enforcement recommendations with the target analysis data, enhancing the transparency and traceability of enforcement decisions. This provides enforcement personnel with references and facilitates their enforcement in current cases. By displaying the main feedback on the enforcement recommendations and supplementing the target analysis data with hidden links, the information presented is clear and concise, avoiding overwhelming enforcement personnel with too much information. This allows them to extract useful information from the feedback and also provides supplementary information to understand why the recommendations were given, offering sufficient supporting evidence for enforcement actions.

[0076] In one embodiment of the present invention, the suggestion feedback module further includes: an interactive query unit; The interactive query unit is used to respond to law enforcement officers' requests for further information regarding the feedback displayed information.

[0077] The further information query request is based on the information displayed in the feedback display unit, and involves further queries, such as: opening hidden links, obtaining more historical cases similar to the representative historical cases, etc. When responding to an information query request, the system analyzes the target query information to determine if it matches the information in the target analysis data. If it does, the target query information is retrieved directly from the target analysis data, and feedback is provided. If not, a secondary related information retrieval is performed by the retrieval analysis module. Based on the information query request, the target query is retrieved from the knowledge base using retrieval enhancement generation technology, and the suggestion feedback module then provides feedback based on the target query information.

[0078] The aforementioned technical solution, through an interactive query unit, enables law enforcement officers to directly conduct further inquiries based on the displayed feedback information. This facilitates follow-up questions based on the feedback information, providing convenience for law enforcement officers and enabling the administrative law enforcement auxiliary system to offer them more reference information for law enforcement, thus providing assurance for their law enforcement activities. Furthermore, it allows for customized information responses based on the needs of law enforcement officers during further inquiries, avoiding the time wasted by re-entering case information into the administrative law enforcement auxiliary system. This also meets the needs of law enforcement officers, improving the operational effectiveness and user satisfaction of the administrative law enforcement auxiliary system.

[0079] Those skilled in the art should understand that the first, second, and third in this invention merely refer to different application stages.

[0080] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0081] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. An administrative law enforcement auxiliary system based on a large language model and retrieval-enhanced generation technology, characterized in that, include: The analysis and processing module is used to acquire case information from law enforcement scenarios and to understand and analyze the case information through a large language model to obtain case description information. The retrieval and analysis module is used to retrieve related information from the knowledge base based on the case description information and retrieval enhancement generation technology to obtain target analysis data. The suggestion feedback module is used to determine law enforcement recommendations based on target analysis data, and then display the law enforcement recommendations in conjunction with the target analysis data.

2. The administrative law enforcement auxiliary system according to claim 1, characterized in that, The analysis and processing module includes: The acquisition unit is used to acquire case information for law enforcement scenarios; The processing unit is used to understand and analyze case information using a large language model, including: preprocessing the case information to obtain first-processed information; and using a standardized processing model to vectorize the first-processed information in combination with the context to obtain case description information.

3. The administrative law enforcement auxiliary system according to claim 2, characterized in that, The acquisition unit also acquires characteristic information of law enforcement personnel, thereby optimizing the case information by combining the characteristic information of law enforcement personnel before the processing unit performs information preprocessing on the case information.

4. The administrative law enforcement auxiliary system according to claim 2, characterized in that, Before the processing unit combines the large language model with the case information for understanding and analysis, it fine-tunes the large language model according to the field of administrative law enforcement.

5. The administrative law enforcement auxiliary system according to claim 1, characterized in that, The retrieval and analysis module includes: The retrieval and analysis unit is used to configure the indexing mechanism for the knowledge base and perform associated indexing in the corresponding knowledge base according to the case description information based on the matching indexing mechanism to obtain the target associated information; The results generation unit is used to enhance the generation of search results based on target association information to obtain target analysis data.

6. The administrative law enforcement auxiliary system according to claim 5, characterized in that, The associated information includes: first associated information, second associated information, and third associated information; When the result generation unit enhances the generation of search results based on target association information, it organizes the first association information in conjunction with the third association information to obtain the first target analysis data, and organizes it based on the second association information to obtain the second target analysis data.

7. The administrative law enforcement auxiliary system according to claim 5, characterized in that, The retrieval and analysis module includes: an update management unit; The update management unit is used to dynamically update the knowledge base, and at the same time, to manage the version of the knowledge base as the version is updated.

8. The administrative law enforcement auxiliary system according to claim 7, characterized in that, The knowledge base includes: a legal and regulatory database, a historical case database, and a policy document database. When the update management unit performs dynamic knowledge updates to the knowledge base, it also performs regular updates to the historical case database. The legal and regulatory database and the policy document database are monitored for adjustments, and the updated version is determined based on the adjustment monitoring information. Then, the updated version is scheduled for a specific time based on the implementation time.

9. The administrative law enforcement auxiliary system according to claim 6, characterized in that, The suggestion feedback module includes: The suggestion generation unit is used to determine law enforcement recommendations based on the analysis of the primary objective data. The feedback display unit is used to display the law enforcement recommendations in conjunction with the target analysis data. It displays the main feedback for the law enforcement recommendations and additional feedback for the target analysis data through hidden links.

10. The administrative law enforcement auxiliary system according to claim 9, characterized in that, The suggestion feedback module also includes: an interactive query unit; The interactive query unit is used to respond to law enforcement officers' requests for further information regarding the feedback displayed information.

Images