Regulation consultation method, readable storage medium and program product

Abstract

The invention provides a regulation consultation method, a readable storage medium and a program product, and the method is characterized in that the method comprises the steps: obtaining a regulation consultation request for a target driving scene, and the regulation consultation request comprises consultation problem information and driving background information; inputting the question information and the driving background information into an intention recognition model, and obtaining a consultation intention output after the intention recognition analysis; and inputting the consultation intention into a law and regulation suggestion model, and obtaining driving-related law and regulation suggestions which meet the consultation intention and are output after analysis of the law and regulation suggestion model.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and more particularly to a method for consulting regulations, a readable storage medium, and a program product. Background Technology

[0002] As laws and regulations develop year by year, their quantity and complexity also increase. Traditional manual management models face multiple bottlenecks: manual retrieval is time-consuming and inefficient, making it difficult to accurately locate the required clauses; the applicability judgment of regulations is easily affected by subjective experience, especially when dealing with complex clauses and a large number of cases, resulting in insufficient accuracy; regulations are updated frequently, and manual follow-up is lagging behind, which can easily lead to information gaps; in addition, the ability to conduct in-depth correlation analysis and cross-regulatory reasoning is lacking, making it difficult to support the decision-making needs of complex scenarios. Therefore, the industry urgently needs to improve the quality and efficiency of the regulatory management process.

[0003] In related technologies, intelligent regulatory management systems are commonly used to perform keyword matching and retrieval of laws and regulations based on user-input terms. However, this method still has significant drawbacks: retrieval tools that rely on keyword matching cannot interpret the user's true intent, leading to biased results. At the same time, the system cannot fully realize semantic understanding and correlation analysis capabilities, making it difficult to cope with complex consultation scenarios involving multiple regulations and conditions, and thus failing to meet the needs of precise and dynamic regulatory management. Summary of the Invention

[0004] In view of this, the present invention provides a regulatory consultation method, a readable storage medium, and a program product to address the shortcomings of related technologies.

[0005] Specifically, this specification is implemented through the following technical solution:

[0006] According to a first aspect of this specification, a method for consulting on regulations is provided, the method comprising:

[0007] Obtain regulatory consultation requests for the target driving scenario, wherein the regulatory consultation requests include consultation question information and driving background information;

[0008] The question information and the driving background information are input into the intent recognition model, and the consultation intent output after the intent recognition analysis is obtained;

[0009] The consultation intent is input into the regulatory recommendation model, and the driving-related regulatory recommendations that conform to the consultation intent are obtained after the regulatory recommendation model is analyzed.

[0010] According to a second aspect of this specification, a regulatory consultation device is provided, the device comprising:

[0011] The consultation request acquisition unit is used to acquire regulatory consultation requests for a target driving scenario. The regulatory consultation request includes consultation question information and driving background information.

[0012] The consultation intent acquisition unit is used to input the question information and the driving background information into the intent recognition model, and to acquire the consultation intent output after the intent recognition analysis;

[0013] The regulatory advice acquisition unit is used to input the consultation intent into the regulatory advice model and obtain driving-related regulatory advice that conforms to the consultation intent after analysis by the regulatory advice model.

[0014] According to a third aspect of this specification, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0015] According to a fourth aspect of this specification, a computer program product includes a computer program / instructions that, when executed by a processor, implement the steps of the method described in the first aspect.

[0016] This manual describes a method for obtaining regulatory consultation requests containing both the consultation question and driving background information. These requests are then sequentially input into an intent recognition model and a regulatory suggestion model for processing. This effectively overcomes the shortcomings of relying on keyword matching. The intent recognition model integrates the question with driving background information specific to the vehicle's driving scenario, accurately interpreting the user's true consultation intent and avoiding result deviations caused by semantic ambiguity or missing context. Simultaneously, the regulatory suggestion model performs deep semantic understanding and correlation analysis based on the consultation intent, outputting matching driving-related regulatory suggestions. This significantly improves the ability to handle complex scenarios involving multiple regulations and conditions in driving situations, ensuring the accuracy and applicability of the suggestions. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the present invention, the accompanying drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0018] Figure 1 This is a schematic diagram of the architecture of a regulatory-based consulting system shown in an embodiment of the present invention;

[0019] Figure 2 This is a flowchart illustrating a regulatory consultation method according to an embodiment of the present invention;

[0020] Figure 3This is a flowchart illustrating another regulatory consultation method according to an embodiment of the present invention;

[0021] Figure 4 This is a schematic structural diagram of an electronic device according to an embodiment of the present invention;

[0022] Figure 5 This is a block diagram illustrating a regulatory consultation device according to an embodiment of the present invention. Detailed Implementation

[0023] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the present invention.

[0024] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular forms “a,” “the,” and “the” used in this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0025] It should be understood that although the terms first, second, third, etc., may be used in this invention to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first information may also be referred to as second information without departing from the scope of this invention, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0026] Figure 1 This is a schematic diagram of the architecture of a regulatory consultation system provided in an exemplary embodiment. Figure 1 As shown, the system may include a regulatory consultation terminal 12 and a regulatory consultation platform 14.

[0027] The regulatory consultation terminal 12 is a front-end device that initiates regulatory consultation requests and can be deployed in vehicles, mobile terminals, or driving management systems. This terminal, through the integration of sensors, a human-machine interface, and a communication module, can collect consultation question information and driving background information for the target driving scenario in real time, and send structured request data to the regulatory consultation platform 14 based on network communication protocols. The terminal supports multimodal input and data preprocessing to ensure the integrity of the consultation request and transmission efficiency, while providing a visual interface for users to view the final regulatory recommendations, forming a consultation closed loop. It is understood that the above can also be used independently as electronic devices such as PCs, mobile phones, tablets, laptops, PDAs (Personal Digital Assistants), wearable devices (such as smart glasses, smartwatches, etc.), etc., and one or more embodiments in this specification do not limit this.

[0028] Server 11 can be a physical server containing an independent host, or it can be a virtual server hosted in a host cluster. During operation, server 11 can run server-side programs for a certain application to implement the relevant functions of that application. For example, when server 11 runs a program for service xx, it can act as a corresponding service platform for xx.

[0029] The regulatory consultation platform 14 is the core execution component of the aforementioned system. It can be a physical server containing an independent host, or the server 11 can be a virtual server hosted in a host cluster. It has a built-in intent recognition model and a regulatory suggestion model. After receiving a consultation request from a terminal, the platform first performs joint semantic analysis of the question information and background information through the intent recognition model, and uses the contextual understanding capabilities of a large model to identify the user's implicit needs, outputting a precise consultation intent. Subsequently, the regulatory suggestion model performs cross-regulatory reasoning based on intent association with a multi-source regulatory library, combined with timely clauses and historical precedents, to generate customized suggestions adapted to the driving scenario. Through a model-layered collaboration and real-time data update mechanism, the platform solves the problems of intent misjudgment and shallow regulatory association in traditional systems, achieving high-precision and dynamic regulatory consultation services.

[0030] The following describes in detail, with reference to the accompanying drawings, embodiments of the regulatory consultation method described in this specification.

[0031] Figure 2 This is a flowchart illustrating an exemplary embodiment of a regulatory consultation method as shown in this specification. Figure 2 As shown, the method may include the following steps:

[0032] Step S202: Obtain a regulatory consultation request for the target driving scenario. The regulatory consultation request includes consultation question information and driving background information.

[0033] This method first obtains regulatory consultation requests for the target driving scenario. These requests contain two core types of information: first, the user's consultation questions, such as "Is lane changing in autonomous driving mode compliant?" or "What responsibility do I bear for rear-ending other vehicles?", typically represented in natural language text or voice; second, driving background information associated with the driving scenario, including but not limited to basic vehicle information such as vehicle model, license plate number, and annual inspection status; vehicle driving information such as speed, driving mode, road type, weather conditions, country / region of origin, geographical location, and historical consultation information associated with the vehicle. Specifically, in cases where a vehicle accident is detected, additional vehicle accident information describing vehicle damage and accident scene characteristics can be added to the aforementioned driving background information. Besides text and voice, this driving background information can also be represented in multimedia formats such as images and videos. By integrating multi-dimensional driving scenario data, this driving background information can support accurate semantic analysis and personalized regulatory recommendations. After being structured and encapsulated, the request data is transmitted over the network to the regulatory consultation platform, providing a multi-dimensional input basis for subsequent analysis. In addition, the so-called target driving scenario can cover every scenario when the user drives the vehicle by default, or it can be set to specific scenarios such as highway driving or parking according to actual needs.

[0034] Regarding the request for legal consultation itself, it can be triggered based on different modes and obtained by the legal consultation platform.

[0035] In one embodiment, a user can initiate a regulatory consultation request by performing a consultation operation in the aforementioned target driving scenario. Specifically, the user's consultation request in the target driving scenario can be obtained through the human-machine interface (such as a touch screen or voice assistant) of an in-vehicle terminal or mobile device. This consultation operation includes, but is not limited to: the user inputting a natural language question, such as asking "Is it permissible for autonomous driving to change lanes on the current road segment?", clicking a preset consultation tag on the interface, such as "Accident Liability Determination" or "Regional Speed ​​Limit Inquiry", or triggering an emergency consultation button, such as clicking "One-Click Legal Support" in an accident scenario. In this embodiment, the system can simultaneously collect consultation question information determined based on the consultation operation and real-time driving background information to form a structured regulatory consultation request. It should be emphasized that the aforementioned driving background information does not need to be actively provided by the user, but is determined by the system actively accessing information maintained in relevant memory or sensors while acquiring the consultation question information. This specification does not limit this.

[0036] In another embodiment, when the vehicle meets preset consultation conditions in the target driving scenario, it can proactively initiate a regulatory consultation request. These preset conditions may include: 1. Triggered by a risk scenario: high-risk events such as collisions, emergency braking, and lane departure violations are detected by onboard sensors or environmental perception modules; 2. Triggered by compliance verification: conflicts are determined based on the vehicle's real-time driving behavior and the locally stored behavior database. For example, if speeding or driving behavior in autonomous driving mode conflicts with the current speed exceeding the road speed limit in the aforementioned behavior database, a regulatory consultation request is initiated; 3. Triggered by cross-regional driving: when the vehicle enters a new administrative jurisdiction, a regional regulatory compatibility query can be automatically initiated. In summary, in this mode, the system can automatically extract driving background information, such as vehicle damage data at the time of an accident, vehicle speed at the time of speeding, and road speed limit signs, and generate default consultation questions, such as "autonomous driving compliance requirements for cross-border driving," without requiring manual input from the user.

[0037] As can be seen, the data collected by the two modes above are transmitted to the regulatory consultation platform after being standardized and packaged, ensuring two-way coverage of proactive human consultation and intelligent system early warning, and improving the real-time performance and scenario adaptability of regulatory services.

[0038] Step S204: Input the question information and the driving background information into the intent recognition model, and obtain the consultation intent output after the intent recognition analysis.

[0039] After receiving a request, the regulatory consultation platform can invoke an intent recognition model to perform joint semantic analysis of the consultation question information and background information. This intent recognition model acts as an intent parsing expert, leveraging the contextual understanding capabilities of a pre-trained large model and considering the spatiotemporal characteristics of driving scenarios to identify the user's implicit needs and eliminate ambiguity, such as distinguishing between "lane change compliance" and "lane change liability determination," ultimately outputting a structured consultation intent. This intent can be based on standardized labels or vectorized feature representations, such as "regulatory applicability analysis of cross-lane changes in L3 autonomous driving scenarios," thereby clarifying the core objectives and constraints of regulatory retrieval and laying the foundation for accurate regulatory matching.

[0040] The process of analyzing question information and driving background information by the above-mentioned intent recognition model can be realized through a multi-agent collaborative architecture. The process of obtaining the consultation intent by analyzing the question information and driving background information can be broken down into multiple steps, such as question parsing and entity recognition, background fusion and scene modeling, intent disambiguation and weight calculation, and intent generation and standardized output. The above-mentioned intent recognition model includes multiple agents that correspond one-to-one with multiple virtual experts, and each step can be executed by the corresponding agent among the multiple agents.

[0041] For example: Regarding the process of problem parsing and entity recognition, the first intelligent agent, acting as a semantic parsing expert, can perform natural language understanding on the consultation question information, extract key entities such as "autonomous driving" and "lane change," and operation objects such as "vehicle" and "driver," while identifying the question type; Regarding the process of background fusion and scene modeling, the second intelligent agent, acting as a scene modeling expert, can construct a dynamic driving scene graph based on vehicle status, geographical location, and environmental data in the driving background information, and spatiotemporally associate it with the question entities; Regarding the process of intent disambiguation and weight calculation, the third intelligent agent, acting as an intent reasoning expert, can associate the voice question with accident scene images through multimodal data alignment to eliminate semantic ambiguity, and calculate the confidence weight of different intent hypotheses by combining a preset rule base and a historical case base; Regarding the process of intent generation and standardized output, the fourth intelligent agent, acting as an intent standardization expert, can generate a structured consultation intent based on the weight ranking results, such as "L3 autonomous driving lane change compliance determination on a rainy highway," and encode it into an intent label or vectorized feature that the platform can recognize, for use by the regulatory suggestion model. In summary, the first intelligent agent focuses on natural language processing, the second intelligent agent specializes in multi-source data fusion, the third intelligent agent strengthens logical reasoning, and the fourth intelligent agent ensures output standardization. Through a dynamic routing mechanism, each intelligent agent can adaptively allocate computing resources based on the type and complexity of the input data. For example, in accident scenarios, the second and third intelligent agents are activated first, achieving a dual improvement in efficient collaboration and intent generation accuracy.

[0042] Understandably, the steps described above can be adjusted in detail according to actual needs, resulting in a process with more or fewer steps, and different steps can be executed by the same or different agents.

[0043] Step S206: Input the consultation intent into the regulatory suggestion model, and obtain the driving-related regulatory suggestions that conform to the consultation intent after analysis by the regulatory suggestion model.

[0044] Based on the generated consultation intent, the regulatory recommendation model can generate customized regulatory recommendations through semantic relevance calculation and logical reasoning. The recommendations can include specific regulatory entries, explanations of applicable conditions, relevant case analyses, and implementation suggestions, and are fed back to the user through a visual interface on the terminal, forming a closed-loop process from intent recognition to decision support. It is worth noting that the aforementioned intent recognition models can constitute different sub-models of the integrated model, such as the first and second sub-models in a Mixture of Experts (MoE) model, or they can exist as independent models; this specification does not impose any restrictions on this.

[0045] In the process of providing driving-related regulatory advice, the aforementioned regulatory advice model actually needs to select and generate customized regulatory advice from the target knowledge base based on the consultation intent output by the intent recognition model, thereby transforming the regulatory clauses into action guidelines with clear conditions, reducing the user's understanding threshold while ensuring that the above advice is adapted to the needs of driving scenarios.

[0046] In one embodiment, the regulatory compliance platform can obtain driving-related regulatory suggestions generated by the regulatory suggestion model after filtering target regulatory content that matches the consultation intent from the target knowledge base. Specifically, based on the consultation intent, relevant regulatory entries can first be retrieved from the target knowledge base, which consists of multi-source heterogeneous data, including: a basic regulatory base, namely national / regional level general traffic regulations and autonomous driving management specifications; a local regulatory base, namely provincial / municipal level special road regulations and temporary traffic control regulations; a case base, namely historical case law rules for liability determination and court discretionary bias data; and a time-sensitive rule base, namely dynamically updated information such as regulatory version number, effective date, and repeal status. During the screening process, the regulatory recommendation model calculates semantic similarity through clause embedding matching, for example, based on Bidirectional Encoder Representations from Transformers (BERT), and performs logical constraint verification based on conditions such as region, vehicle type, and driving mode to determine a preliminary set of candidate regulations. This set of candidate regulations, as the target regulatory content, can be transformed into user-understandable recommendation text. The final recommendation is then fed back through the regulatory consultation terminal in a combination of text and graphics, and supports voice broadcasting and interactive Q&A.

[0047] It's worth noting that the aforementioned set of candidate regulations can be used directly as target regulations, or a multi-dimensional weighting evaluation can be conducted to determine the target regulations. For example, regional adaptability can be considered, prioritizing regulations governing the vehicle's current location over general provisions; timeliness can be considered, prioritizing the latest effective version and excluding repealed or suspended provisions; or case support can be considered, prioritizing frequently adopted case law to improve the enforceability of the recommendations. Of course, for conflicting provisions, dynamic compliance strategies can be generated by combining driving route planning information from the driving background. For example, if different regions have conflicting regulations for the same driving behavior, a dynamic compliance strategy such as "Current area prohibits lane changes; it is recommended to comply when entering the X provincial border 3 kilometers later" can be generated based on the expected time of entry into the next administrative region.

[0048] Furthermore, the construction of the aforementioned target knowledge base employs a multi-faceted data collection strategy, primarily encompassing two core approaches: automated online data collection and user-initiated uploads. Regarding online data collection, compliant and efficient methods are employed to collect data from publicly accessible websites that meet the collection criteria, such as government announcement platforms, industry information portals, and academic resource repositories. Natural language processing technology is used to automatically parse, structure, and quality-filter webpage content. For user uploads, the system provides standardized data upload interfaces, supporting submissions of text, documents, images, and other formats. A data review mechanism and metadata annotation standards are also included to ensure the accuracy, compliance, and domain suitability of user-contributed data. The collaborative operation of these two collection methods enables dynamic updates and scalable expansion of the knowledge base content, meeting the knowledge retrieval and semantic analysis needs of various application scenarios.

[0049] This manual also supports dynamic configuration of newly added regulatory content to ensure the timeliness and completeness of the target knowledge base.

[0050] In one embodiment, the regulatory consultation platform can obtain a regulatory configuration request targeting a knowledge base and containing regulatory content to be configured. Based on this content, it generates classification prompts and inputs them into a large-scale classification model. The model then analyzes the data and stores the regulatory content to be configured and the corresponding classification results in the target knowledge base. Specifically, the regulatory content to be configured can be newly added regulatory text, revised clauses, or user-defined rules, and its data format includes text, PDF documents, or structured data tables. The regulatory content can be preprocessed to extract key features such as the regulatory title, issuing authority, effective date, and applicable region, and classification prompts can be generated based on predefined classification rules. For example, "You are a regulatory classification assistant and need to classify the user-inputted automotive-related regulatory content into corresponding automotive-related fields; the fields are: {{function_str}}. Please only output the field names to which the classification is applied, requiring complete and accurate classification into the corresponding fields, with a maximum of two. If no field is specified, please output 'None'." Here, the variable `function_str` can be a pre-stored list of fields. In summary, after inputting classification prompts into a large-scale classification model, such as one based on GPT-4, the model can use semantic understanding and contextual association to output regulatory type labels, association levels, and validity indicators that meet the requirements of the classification prompts. The regulatory content to be configured is then associated with its classification results and stored in the target knowledge base, with an index created for subsequent retrieval by the regulatory suggestion model. Furthermore, the classification results from the aforementioned large-scale classification model can be further compared and filtered by other large-scale classification models, thus contributing to the accurate analysis of the classification results. Specifically, the regulatory content to be configured and its corresponding classification results can be input into a different validation model than the large-scale classification model. This validation model can be functionally identical to the large-scale classification model, and the final classification result is set as the overlapping portion of the validation and classification results of the validation model. This means that the current classification result simultaneously conforms to the classification results independently output by both the large-scale classification model and the validation model.

[0051] Based on the classification results output by the aforementioned classification model, a semantic network between regulations can be further constructed through the association model to enhance the association retrieval capability of the target knowledge base.

[0052] In one embodiment, the regulatory consultation platform can generate relevant prompts based on the regulatory content to be configured, input these prompts into a large-scale correlation model, and store the knowledge graph of the regulatory content to be configured, output by the large-scale correlation model, into the target knowledge base. Specifically, based on the regulatory type tags output by the large-scale classification model and the regulatory content to be configured, prompts required for correlation analysis can be extracted, which may include:

[0053] 1. Entity keywords: regulatory subject, such as "autonomous driving operator"; behavior object, such as "lane change" and "speed limit"; constraints, such as "rainy day" and "highway";

[0054] 2. Relationship trigger words: Reference relationship between laws and regulations, such as "according to Article Y of Law X", conflict markers, such as "inconsistent with Regulation Z", complementary descriptions, such as "must be implemented in conjunction with Standard A", etc.

[0055] 3. Spatiotemporal attributes: effective date of regulations, expiration date, applicable geofence.

[0056] Of course, the operation of extracting relevant keywords can also be handled by a large model, and the corresponding keyword could be: "Task": "Entity and Relation Extraction".

[0057] "Dataset":"Regulatory Related Information",

[0058] "instruction": "Extract entities and relationships from the following regulatory text. Entity types include regulatory name, clause, subject, etc., and relationship types include reference, application, interpretation, etc. Please answer in the following format:\nEntity: [Entity Name] (Entity Type)\nRelationship: [Entity 1] (Relationship Type) [Entity 2]" This instruction does not impose any restrictions on this.

[0059] In summary, after inputting the identified association prompts into a large association model, such as one based on a GAT graph attention network, the model can perform the following operations: First, extract entity relationships, that is, identify the types of associations between the regulations to be configured and existing regulations and cases in the knowledge base, and perform conflict detection and annotation to compare clause content, automatically mark potential conflicts, and record the basis for conflict resolution; finally, update the corresponding knowledge graph, so that the legal entities are nodes and the association relationships are edges, dynamically update the legal knowledge graph in the target knowledge base, and store it in the graph database.

[0060] Based on this, when the regulatory suggestion model actually filters the above-mentioned target regulatory content, in addition to semantic matching, the solution in this manual can additionally call the above-mentioned knowledge graph to perform the following enhanced operations: 1. Related extended retrieval: Based on the graph nodes associated with the consultation intent, automatically expand the retrieval of relevant regulations. For example, when querying "autonomous driving lane change", the general lane change rules in its superior law, the Road Traffic Safety Law, are returned simultaneously; 2. Conflict warning prompt: If the suggested content involves conflicting clauses marked in the graph, a warning is added to the feedback, such as "Note: Local regulations conflict with the speed limit rules of City D you are about to enter"; 3. Multi-regulation joint reasoning: Based on the relationship chain in the graph, deduce compliance strategies under complex scenarios, such as "Cross-border transportation must simultaneously meet the regulations of the origin, transit, and destination regions".

[0061] Similar to the intent recognition model described above, the analysis process of consultation intent in the regulatory recommendation model can be implemented through a multi-agent collaborative architecture. The process by which the regulatory recommendation model analyzes consultation intent to obtain the aforementioned driving-related regulatory recommendations can be broken down into multiple steps, such as initial regulatory screening and semantic matching, conflict detection and prioritization, recommendation generation and executability optimization, and optional compliance verification and feedback correction. Furthermore, the intent recognition model includes multiple agents corresponding one-to-one with multiple virtual experts, and each step can be executed by the corresponding agent among these multiple agents.

[0062] For example, regarding the aforementioned process of initial screening and semantic matching of regulations, a fifth intelligent agent, acting as a regulatory retrieval expert, can perform a coarse-grained retrieval from the target knowledge base based on core tags in the consultation intent, such as "autonomous lane changing." Specifically, this intelligent agent can adopt the following hybrid retrieval strategy: performing similarity calculations between the intent vector and the embedded representation of the regulatory clauses based on, for example, a semantic vector matching algorithm; and excluding obviously inapplicable clauses based on the region, time, and vehicle type in the driving background information, ultimately outputting a preliminary candidate regulatory set. Regarding the aforementioned conflict detection and prioritization process, a sixth intelligent agent, acting as a conflict resolution expert, can conduct in-depth analysis of the candidate regulatory set. Specifically, it can identify contradictions between different regulatory levels, such as conflicts between local regulations and national standards, and assign priorities based on principles such as "priority of higher-level laws" and "newer laws prevail over older laws." Simultaneously, it can combine real-time vehicle status and environmental parameters, such as road slippage coefficients, to calculate the applicability weight of clauses. For example, in rainy scenarios, the weight of special weather regulations is increased by 30%. Furthermore, it can link historical case databases to add confidence bonuses to frequently cited clauses, ultimately outputting a ranked sequence of conflict-free regulations and their applicability scores. Regarding the aforementioned process of generating and optimizing the feasibility of recommendations, a seventh intelligent agent, acting as a recommendation generation expert, can transform the screened regulations into actionable recommendations. Specifically, it can extract mandatory, prohibitive, and authorizing clauses from the regulations and map them as "mandatory operations," "prohibited behaviors," and "optional decisions," respectively. It can also transform abstract legal expressions into specific driving parameters, such as "maintaining a safe following distance" into "distance from the vehicle in front ≥ 50 meters," and verify feasibility by linking real-time data from vehicle sensors. Finally, based on the type of terminal device, it can dynamically generate text summaries, flowcharts, or voice broadcasts. Furthermore, for the optional steps of compliance verification and feedback correction, an eighth intelligent agent, acting as a dynamic verification expert, can provide real-time monitoring during the recommendation execution phase. For example, it can obtain vehicle control signals such as steering angle and acceleration through the interface of the On-Board Diagnostics (OBD) system to verify whether driving behavior meets the recommendation requirements; or, if the regulations are updated after recommendation generation, it can automatically trigger a re-evaluation of the recommendations and push a revision notification.

[0063] The following is combined Figure 3 The flowchart for another regulatory-based consultation method is illustrated, such as... Figure 3 As shown, the method includes the following steps.

[0064] Step S302: The regulatory consultation terminal initiates a regulatory consultation request.

[0065] In one embodiment, the regulatory consultation terminal can trigger consultation requests through user-initiated operation or automatic system detection. The terminal integrates onboard sensors and a human-machine interface module to collect question and background information in real time and encapsulate it into structured request data. In automatic trigger mode, the system generates default consultation questions based on preset rules. For example, when a Level 4 autonomous vehicle enters the border of country B from country A, the Global Positioning System (GPS) triggers geofencing rules and automatically generates a consultation request. This request includes the consultation question "Compliance requirements for lane changing with autonomous driving within country B," and simultaneously collects the vehicle's location (the border between countries A and B), driving mode (Level 4), and road type (highway) as driving background information.

[0066] Step S304: The first sub-model analyzes the consultation intent.

[0067] In one embodiment, it is assumed that the intent recognition model constitutes the first sub-model in the expert hybrid model, and the regulatory recommendation model constitutes the second sub-model in the expert hybrid model. The first sub-model can adopt a multi-agent collaborative architecture to jointly analyze the problem and background information using the following steps: 1. Semantic parsing experts extract key entities; 2. Scene modeling experts construct a spatiotemporal correlation graph; 3. Intent reasoning experts eliminate ambiguity and calculate intent weights; 4. Standardization experts generate structured intent labels. For example, given the input consultation question information "cross-border autonomous driving lane change rules" and the aforementioned driving background information, the output intent label is "L4 level autonomous driving cross-border lane change compliance determination, region: country B; road type: highway".

[0068] Step S306: The second sub-model retrieves regulations and invokes the knowledge graph.

[0069] In one embodiment, the regulatory retrieval expert of the second sub-model described above can perform semantic matching and conditional filtering from the target knowledge base based on intent tags and BERT embedding similarity calculation to generate a candidate regulatory set. Simultaneously, it invokes the knowledge graph for associated extended retrieval and conflict detection. For example, if the retrieval result yields Article 5 (territorial priority) of the "Regulations on the Management of Automated Driving in Country B" and Chapter 12 (supplementary application) of the "International Cross-border Vehicle Agreement," the knowledge graph marks the speed limit rules of the two as conflicting.

[0070] Step S308: Conflict resolution and dynamic strategy generation in the second sub-model.

[0071] In one embodiment, the conflict resolution expert of the second sub-model calculates the priority of clauses based on hierarchical rules and scenario parameters, generating a dynamic compliance strategy. Finally, the suggestion generation expert transforms the regulations into actionable actions and multimodal content. For example, given the priority given to country B's territory, a suggestion could be generated to "activate the right lane changing mode after entering country B's territory for 500 meters," and the clause "lane changing requires activating the regional compliance mode" could be linked to the vehicle's turn signal.

[0072] Step S310: The second sub-model verifies compliance and provides real-time feedback.

[0073] In one embodiment, the dynamic verification expert in the second sub-model can monitor driving behavior through the OBD interface to verify whether it meets the recommended requirements. If a regulatory update or behavioral deviation is detected, a real-time alarm and suggested revision are triggered. For example, if the vehicle does not activate the compliance mode when changing lanes, the system provides a voice prompt "Please switch to the right lane lane changing mode immediately" as a driving-related regulatory suggestion, while the OBD log records the violation.

[0074] Step S312: Display driving-related regulatory suggestions to the regulatory consultation terminal.

[0075] In one embodiment, the terminal presents the aforementioned driving-related regulatory recommendations, conflict warnings, and legal consequences through a visual interface, while also supporting user rating feedback. Based on feedback data and consultation records, the system automatically updates the case library and knowledge graph weights to optimize subsequent search accuracy. For example, if the interface displays "Lane changing is only permitted in the right lane within Country B (according to Article 5)," and the user clicks "Valid" to rate the scenario, the case library adds applicable records for that scenario, increasing the search priority for similar consultations.

[0076] Figure 4 This is a schematic structural diagram of an electronic device according to an exemplary embodiment. Please refer to... Figure 4 At the hardware level, the electronic device includes a processor 402, an internal bus 410, a network interface 404, memory 406, and non-volatile memory 408, and may also include other necessary hardware. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it, forming a risk code detection device at the logical level. Of course, in addition to software implementation, this specification does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution subject of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0077] Figure 5 This invention illustrates a block diagram of a regulatory consultation device according to an embodiment of the invention. Please refer to... Figure 5 This device can be applied to, for example Figure 4The device shown, in order to implement the technical solution described in this invention, includes:

[0078] Consultation request acquisition unit 502 is used to acquire regulatory consultation requests for a target driving scenario, wherein the regulatory consultation request includes consultation question information and driving background information;

[0079] The consultation intent acquisition unit 504 is used to input the question information and the driving background information into the intent recognition model, and to acquire the consultation intent output after the intent recognition analysis;

[0080] The regulatory advice acquisition unit 506 is used to input the consultation intent into the regulatory advice model and acquire driving-related regulatory advice that conforms to the consultation intent after analysis by the regulatory advice model.

[0081] Optionally, the consultation request acquisition unit 502 is specifically used for:

[0082] Obtain the regulatory consultation request initiated by the user when performing a consultation operation in the target driving scenario; or,

[0083] Obtain regulatory consultation requests initiated by the vehicle when the preset consultation conditions are met in the target driving scenario.

[0084] Optionally, the regulatory recommendation acquisition unit 506 is specifically used for:

[0085] The regulatory suggestion model obtains driving-related regulatory suggestions based on the target regulatory content generated after filtering target regulatory content from the target knowledge base according to the consultation intent.

[0086] Optionally, the device further includes:

[0087] The regulatory configuration unit is used to obtain a regulatory configuration request for the target knowledge base, wherein the regulatory configuration request contains the regulatory content to be configured.

[0088] The regulatory classification unit is used to generate classification prompts based on the regulatory content to be configured, input the classification prompts into the classification model, and store the regulatory content to be configured and the corresponding classification results output by the classification model after analysis into the target knowledge base.

[0089] Optionally, the device further includes:

[0090] The classification result verification unit is used to input the regulatory content to be configured and the corresponding classification result into a verification model that is different from the classification model, and to set the classification result as the overlapping part of the verification result of the verification model and the classification result.

[0091] Optionally, the device further includes:

[0092] The knowledge graph storage unit is used to generate associated prompt words based on the content of the regulations to be configured, input the associated prompt words into the association model, and store the knowledge graph of the content of the regulations to be configured, which is output by the association model after analysis, into the target knowledge base.

[0093] Optionally, the process by which the intent recognition model analyzes the question information and the driving background information to obtain the consultation intent includes multiple steps, and the intent recognition model includes multiple intelligent agents corresponding one-to-one with multiple virtual experts, wherein each step is executed by the corresponding intelligent agent among the multiple intelligent agents; and / or,

[0094] The process by which the regulatory recommendation model analyzes the consultation intent to obtain the driving-related regulatory recommendations includes multiple steps, and the regulatory recommendation model includes multiple intelligent agents that correspond one-to-one with multiple virtual experts, wherein each step is executed by the corresponding intelligent agent among the multiple intelligent agents.

[0095] Optionally, if the process of the intent recognition model analyzing the question information and the driving background information to obtain the consultation intent includes multiple steps, the process includes the following steps: question parsing and entity recognition, background fusion and scene modeling, intent disambiguation and weight calculation, intent generation and standardized output;

[0096] In cases where the process of obtaining driving-related regulatory recommendations by analyzing the consultation intent using the regulatory recommendation model involves multiple steps, the process includes the following steps: initial screening and semantic matching of regulations, conflict detection and prioritization, recommendation generation and executability optimization, compliance verification and feedback correction.

[0097] Optionally, the driving background information includes at least one of the following: basic vehicle information, vehicle driving information, vehicle accident information, and historical consultation information of the vehicle user.

[0098] Optionally, the intent recognition model and the regulatory recommendation model may constitute different sub-models of the integrated model, or they may be independent models.

[0099] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0100] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the solution in this specification according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0101] Based on the same concept as the methods described above, this specification also provides an electronic device, including: a processor; a memory for storing processor-executable instructions; wherein the processor performs the steps of the method as described in any of the above embodiments by executing the executable instructions.

[0102] Based on the same concept as the methods described above, this specification also provides a computer-readable storage medium having computer instructions stored thereon that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

[0103] Based on the same concept as the methods described above, this specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

[0104] The embodiments of the subject matter and functional operation described in this specification can be implemented in the following ways: digital electronic circuits, tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or combinations thereof. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier for execution by a data processing apparatus or for controlling the operation of a data processing apparatus. Alternatively or additionally, the program instructions may be encoded on artificially generated propagation signals, such as machine-generated electrical, optical, or electromagnetic signals, which are generated to encode information and transmit it to a suitable receiving device for execution by the data processing apparatus. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or combinations thereof.

[0105] The processing and logic flow described in this specification can be executed by one or more programmable computers that execute one or more computer programs to perform corresponding functions by operating on input data and generating output. The processing and logic flow can also be executed by dedicated logic circuitry—such as FPGAs (Field-Programmable Gate Arrays) or ASICs (Application-Specific Integrated Circuits), and the device can also be implemented as dedicated logic circuitry.

[0106] Suitable computers for executing computer programs include, for example, general-purpose and / or special-purpose microprocessors, or any other type of central processing unit. Typically, the central processing unit receives instructions and data from read-only memory and / or random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, such as disks, magneto-optical disks, or optical disks, or the computer will be operatively coupled to such mass storage devices to receive data from or transfer data to them, or both. However, a computer is not required to have such devices. Furthermore, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a GPS receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few.

[0107] Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, such as semiconductor memory devices (e.g., EPROM, EEPROM, and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CD-ROM and DVD-ROM disks. Processors and memory may be supplemented by or incorporated into dedicated logic circuitry.

[0108] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope of the claims, but rather are primarily intended to describe features of specific embodiments of a particular invention. Certain features described in the various embodiments herein may also be implemented in combination in a single embodiment. Conversely, various features described in a single embodiment may also be implemented separately in various embodiments or in any suitable sub-combination. Furthermore, while features may function in certain combinations as described above and even initially claimed in this way, one or more features from a claimed combination may be removed from that combination in some cases, and a claimed combination may refer to a sub-combination or a variation thereof.

[0109] Similarly, although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0110] Therefore, specific embodiments of the subject matter have been described. Furthermore, the processes depicted in the figures are not necessarily shown in a specific order or sequence to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0111] The above description is merely a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of protection of this specification.

Claims

1. A method for consulting on regulations, characterized in that, The method includes: Obtain regulatory consultation requests for the target driving scenario, wherein the regulatory consultation requests include consultation question information and driving background information; The question information and the driving background information are input into the intent recognition model, and the consultation intent output after the intent recognition analysis is obtained; The consultation intent is input into the regulatory recommendation model, and the driving-related regulatory recommendations that conform to the consultation intent are obtained after the regulatory recommendation model is analyzed.

2. The method according to claim 1, characterized in that, The process of obtaining regulatory consultation requests for the target driving scenario includes: Obtain the regulatory consultation request initiated by the user when performing a consultation operation in the target driving scenario; or, Obtain regulatory consultation requests initiated by the vehicle when the preset consultation conditions are met in the target driving scenario.

3. The method according to claim 1, characterized in that, The process of obtaining driving-related regulatory recommendations that conform to the consultation intent, output by the regulatory recommendation model after analysis, includes: The regulatory suggestion model obtains driving-related regulatory suggestions based on the target regulatory content generated by filtering target regulatory content that matches the consultation intent from the target knowledge base.

4. The method according to claim 3, characterized in that, The method further includes: Obtain a regulatory configuration request for the target knowledge base, wherein the regulatory configuration request contains the regulatory content to be configured; Based on the content of the regulations to be configured, classification prompt words are generated, and the classification prompt words are input into the classification model. The content of the regulations to be configured and the corresponding classification results output by the classification model after analysis are stored in the target knowledge base.

5. The method according to claim 4, characterized in that, The method further includes: The regulatory content to be configured and the corresponding classification result are input into a large verification model that is different from the large classification model, and the classification result is set as the overlapping part of the verification result of the large verification model and the classification result.

6. The method according to claim 4, characterized in that, The method further includes: Based on the content of the regulations to be configured, generate related prompt words, input the related prompt words into the large association model, and store the knowledge graph of the content of the regulations to be configured, which is output by the large association model after analysis, into the target knowledge base.

7. The method according to claim 1, characterized in that, The process by which the intent recognition model analyzes the question information and the driving background information to obtain the consultation intent includes multiple steps, and the intent recognition model includes multiple intelligent agents corresponding one-to-one with multiple virtual experts, wherein each step is executed by the corresponding intelligent agent among the multiple intelligent agents; and / or, The process by which the regulatory recommendation model analyzes the consultation intent to obtain the driving-related regulatory recommendations includes multiple steps, and the regulatory recommendation model includes multiple intelligent agents that correspond one-to-one with multiple virtual experts, wherein each step is executed by the corresponding intelligent agent among the multiple intelligent agents.

8. The method according to claim 6, characterized in that, When the process of obtaining the consultation intent by analyzing the question information and the driving background information by the intent recognition model includes multiple steps, the process includes the following steps: question parsing and entity recognition, background fusion and scene modeling, intent disambiguation and weight calculation, intent generation and standardized output; In cases where the process of obtaining driving-related regulatory recommendations by analyzing the consultation intent using the regulatory recommendation model involves multiple steps, the process includes the following steps: initial screening and semantic matching of regulations, conflict detection and prioritization, recommendation generation and executability optimization, compliance verification and feedback correction.

9. The method according to claim 1, characterized in that, The driving background information includes at least one of the following: basic vehicle information, vehicle driving information, vehicle accident information, and historical inquiry information of the vehicle user.

10. The method according to claim 1, characterized in that, The intent recognition model and the regulatory recommendation model constitute different sub-models of the integrated model, or they are independent models.

11. A regulatory consultation device, characterized in that, The device includes: The consultation request acquisition unit is used to acquire regulatory consultation requests for a target driving scenario. The regulatory consultation request includes consultation question information and driving background information. The consultation intent acquisition unit is used to input the question information and the driving background information into the intent recognition model, and to acquire the consultation intent output by the intent recognition model after analysis. The regulatory advice acquisition unit is used to input the consultation intent into the regulatory advice model and obtain driving-related regulatory advice that conforms to the consultation intent after analysis by the regulatory advice model.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 10.

13. A computer program product, characterized in that, Includes a computer program / instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1-10.

Images