Immersive law education system based on artificial intelligence

Abstract

The invention relates to the technical field of intelligent law education and law reasoning, and discloses a law immersive education system based on artificial intelligence, and the system comprises an input processing module which is configured to carry out the multistage semantic analysis of a law text inputted by a user, and generates a structured law fact representation; the legal knowledge graph construction module is connected with the input processing module and is used for receiving the structured legal facts and constructing and updating a legal knowledge graph; and the automatic theorem proving module is connected with the legal knowledge graph construction module and is used for executing multi-hop reasoning on the legal knowledge graph based on probabilistic reasoning. According to the technical scheme, the self-adaptive optimization module and the dynamic reasoning mechanism are combined, the technical effects of adjusting the priority of the legal rule in real time and automatically optimizing the reasoning path are achieved, the defects that the reasoning process is stiff, the adaptability is poor, and reasoning errors are likely to be caused are overcome, and the intelligent decision-making capacity of the system is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent legal education and legal reasoning, and specifically provides a legal immersive education system based on artificial intelligence. Background Art

[0002] With the continuous development of modern legal education and legal practice, some limitations of the traditional legal learning mode have gradually emerged. The breadth and depth of legal professional knowledge pose great challenges for learners in mastering legal provisions, case analysis, and reasoning skills. At the same time, the diverse case types and complex cross-jurisdictional issues in judicial practice also place higher requirements on legal practitioners. Traditional legal education methods mainly rely on books, lectures, and case discussions, lacking the simulation of real scenarios and interactivity, often unable to fully mobilize the enthusiasm of learners, nor can they provide targeted training.

[0003] Existing legal teaching systems usually rely on text and video tutorials to impart knowledge through traditional classroom teaching methods. For legal reasoning and practical training, the combination of theory and case analysis is usually adopted, but this method has certain limitations. For example, in traditional teaching, legal provisions and cases are often presented in text form, making it difficult to effectively display the complex reasoning paths and multi-dimensional conflict issues in the process of legal application. In addition, although some teaching platforms have introduced functions such as mock trials and virtual laboratories, these systems usually lack immersion and interactivity, making it difficult to fully mobilize the enthusiasm and practical ability of learners.

[0004] The main problem of the existing technology lies in the lack of dynamic adaptability to different case types and legal environments. In traditional systems, the legal reasoning process often relies on fixed rules and rigid teaching content, and cannot be adjusted in real time to cope with complex case variables and the personalized needs of learners. This results in poor flexibility of the reasoning process and insufficient participation of learners. Especially in complex cases and cross-jurisdictional application issues, it is easy to cause limitations in teaching effects and misjudgments in legal applications. Therefore, the existing technology cannot provide an intelligent legal education platform that can be dynamically adjusted according to user needs and case complexity. Summary of the Invention

[0005] Aiming at the deficiencies of the existing technology, the present invention provides a legal immersive education system based on artificial intelligence, which solves the problems of the deficiencies of the existing legal education system in terms of dynamic adaptability of legal reasoning, cross-jurisdictional application ability, and immersive interactive experience.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A legal immersive education system based on artificial intelligence, comprising: An input processing module, configured to perform multi-level semantic analysis on the legal text input by the user and generate a structured legal fact representation; A legal knowledge graph construction module, connected to the input processing module, for receiving structured legal facts and constructing and updating a legal knowledge graph; An automatic theorem proving module, connected to the legal knowledge graph construction module, for performing multi-hop reasoning on the legal knowledge graph based on probabilistic reasoning; A reasoning explanation module, connected to the automatic theorem proving module, for visually displaying the reasoning process using explainable artificial intelligence technology; A cross-jurisdictional conflict resolution module, connected to the legal knowledge graph construction module, for detecting and resolving cross-jurisdictional legal provision conflicts; An adaptive optimization module, respectively connected to the cross-jurisdictional conflict resolution module and the automatic theorem proving module, for dynamically optimizing reasoning parameters and the knowledge graph structure based on user feedback and system performance metrics; A VR teaching terminal, connected to the reasoning explanation module, for generating an immersive teaching environment including a three-dimensional virtual courtroom scene, a dynamic case evolution process, and an interactive teaching interface.

[0007] Preferably, the input processing module includes: A text preprocessing unit, configured to perform word segmentation, part-of-speech tagging, and named entity recognition on the input text; A semantic analysis unit, connected to the text preprocessing unit, configured to perform syntactic analysis and semantic role labeling using a BiLSTM model based on an attention mechanism; A relation extraction unit, connected to the semantic analysis unit, configured to extract legal fact triples based on a predefined legal relation template.

[0008] Preferably, the legal knowledge graph construction module includes: An entity relation encoding layer, configured to perform feature encoding on legal provision nodes using a graph convolutional network; A cross-jurisdictional association layer, connected to the entity relation encoding layer, configured to establish a semantic mapping relationship between legal provisions in different jurisdictions through a contrastive learning algorithm; A dynamic update layer, connected to the cross-jurisdictional association layer, configured to update the knowledge graph in real time to reflect the revision of legal provisions.

[0009] Preferably, the cross-jurisdictional association layer includes: A jurisdiction feature extraction unit, configured to extract the legal provision features of a specific jurisdiction; A semantic alignment unit, connected to the jurisdiction feature extraction unit, configured to minimize the embedding distance of similar legal concepts; A bridging edge generation unit, connected to the semantic alignment unit, is configured to establish cross-jurisdiction association edges.

[0010] Preferably, the automated theorem proving module includes: A predicate logic conversion unit, configured to convert structured legal facts into predicate logic expressions including legal subjects, acts, and objects; A multi-hop reasoning unit, connected to the predicate logic conversion unit, is configured to perform multi-hop reasoning based on the attention mechanism on the legal knowledge graph; An evidence credibility evaluation unit, connected to the multi-hop reasoning unit, is configured to calculate the confidence of each reasoning path based on the Bayesian inference framework.

[0011] Preferably, the reasoning explanation module includes: A reasoning path visualization unit, configured to generate a tree-shaped topology graph including a legal provision citation chain; An influence degree analysis unit, connected to the reasoning path visualization unit, is configured to quantify the contribution degree of each legal element to the reasoning result using SHAP values.

[0012] Preferably, the cross-jurisdiction conflict resolution module includes: A conflict detection unit, configured to identify legal provisions with applicable conflicts; A conflict resolution unit, connected to the conflict detection unit, is configured to calculate the applicability score of provisions based on the semantic similarity of provisions, case citation frequency, and the timeliness of legal provisions.

[0013] Preferably, the adaptive optimization module includes: A reinforcement learning agent, configured to dynamically adjust evidence weight parameters, fact similarity weights, and conflict resolution coefficients according to user feedback using the PPO algorithm; A knowledge graph updater, connected to the reinforcement learning agent, is configured to update the knowledge graph structure in real time.

[0014] Preferably, the VR teaching terminal includes: A gesture recognition unit, configured to capture operation instructions of the user on the virtual legal document; A dynamic case evolution unit, connected to the gesture recognition unit, is configured to change the debate process of the virtual court in real time according to the selected reasoning path by the user; A teaching effect evaluation unit, connected to the dynamic case evolution unit, is configured to calculate the attention distribution of the user on key legal elements through eye tracking data.

[0015] The present invention provides an AI-based immersive legal education system. It has the following beneficial effects: 1. The present invention adopts a technical solution combining an adaptive optimization module with a dynamic reasoning mechanism, achieving the technical effect of real-time adjustment of the priority of legal rules and automatic optimization of the reasoning path. Compared with the technical solution of the prior art that rigidly applies fixed rules and is difficult to cope with complex case variables, it solves the shortcomings of its rigid reasoning process, poor adaptability and easy reasoning errors, and significantly improves the intelligent decision-making ability of the system.

[0016] 2. The present invention introduces a technical solution that deeply integrates immersive VR teaching terminals with intelligent interactive systems, achieving the technical effect of three-dimensional dynamic visualization and real-time interactive experience of legal knowledge and reasoning processes. Different from the technical means of single text reading and passive indoctrination learning in traditional legal teaching, it solves the technical bottlenecks of boring teaching, poor sense of participation and insufficient practical training, making legal learning more immersive and practical.

[0017] 3. The present invention adopts the design scheme of cross-jurisdiction conflict resolution mechanism and multi-source legal knowledge graph modeling technology, achieving the technical effect of automatically identifying jurisdiction conflicts and quickly reconstructing applicable paths. Compared with the problems of confusion in jurisdiction application and difficulty in timely resolution of conflicts in traditional systems, it breaks through the technical bottleneck of relying on manual judgment, which leads to low efficiency and high risk of errors, and greatly enhances the system's cross-jurisdiction applicability.

[0018] 4. The present invention combines the technical architecture of intelligent teaching engine and personalized reasoning training algorithm, and successfully achieves the technical effect of dynamically adjusting the difficulty of training tasks and content based on user capabilities. Traditional teaching systems often lack pertinence and cannot teach students in accordance with their aptitude, resulting in low learning efficiency for users. The present invention effectively solves the shortcomings of this model, which is single, slow to feedback, and lacks personalized training mechanism, so that users at different levels can obtain the optimal learning path. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the overall system architecture of the present invention; Figure 2 A schematic diagram of input processing and knowledge graph construction of the present invention; Figure 3 It is a schematic diagram of the reasoning and teaching interaction of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Please see attached Figures 1-3, an embodiment of the present invention provides an AI-based immersive legal education system, including: The main task of the input processing module is to perform multi-level semantic analysis on the legal text input by the user, convert the legal facts in natural language into a structured legal fact representation, and provide basic data for subsequent legal reasoning and knowledge graph construction. This module uses natural language processing (NLP) technology, combined with legal domain-specific terms, to perform text preprocessing, semantic understanding, and relationship extraction.

[0022] 1. Text preprocessing unit This unit first performs basic text processing on the input legal text to ensure efficient subsequent analysis. It includes: Word segmentation: Divide long sentences into word units. For legal texts, a domain-specific word segmentation dictionary is used to handle the segmentation of legal proper nouns and phrases. For example, legal terms such as "tort liability" and "presumption of innocence" need to be segmented according to the special requirements of the legal domain.

[0023] Part-of-speech tagging: Tag the part of speech of each word to determine its grammatical role in the sentence, such as noun, verb, adjective, etc. Here, a BERT-based part-of-speech tagging model is used, which can handle complex sentence patterns and context changes.

[0024] Named entity recognition: This module automatically identifies key information such as legal subjects, victims, and legal provisions in the text. A pre-trained NER model on legal corpus, such as LawBERT, is used to identify proper nouns and entities through a deep neural network combined with character-level information. The NER model uses BERT-based named entity recognition technology to accurately identify entities such as names, locations, and case numbers in legal texts. To adapt to the particularity of the legal domain, a domain-specific NER dataset is introduced and fine-tuned to improve the model's ability to identify legal entities.

[0025] 2. Semantic analysis unit This unit is mainly responsible for converting the grammatical structure of the text into a semantic structure to obtain an abstract representation of legal facts. In terms of technical implementation, a bidirectional LSTM (BiLSTM) model based on BiLSTM and attention mechanism is used and processed through the following steps: Syntactic analysis: The text is parsed into a syntactic tree through BiLSTM to analyze the grammatical relationships in the sentence. For example, in the sentence "Zhang San intentionally injured Li Si", the model can determine that "Zhang San" is the subject, "intentionally injured" is the predicate, and "Li Si" is the object. BiLSTM can consider the context information before and after while processing the text, effectively capturing the grammatical dependencies in the syntax. Combined with the attention mechanism, the model can dynamically focus on the key information in the sentence. In legal texts, certain specific words or phrases are crucial for understanding legal facts, and the attention mechanism can assign higher weights to these important parts.

[0026] Semantic Role Labeling (SRL): Further label the roles of each component in the sentence to identify semantic roles such as the agent, action, object, time, and place. For example, "Zhang San" is the agent, "intentionally injured" is the action, and "Li Si" is the victim. This step is trained jointly by Bidirectional LSTM (BiLSTM) + attention mechanism to improve the ability to capture complex semantics.

[0027] 3. Relationship Extraction Unit The task of this unit is to extract the triples of legal facts, namely the subject, action, and object, from the annotated text. This is achieved through the following techniques: Template-based Relationship Extraction: Use predefined legal relationship templates, such as the form "<subject> <action> <object>", to extract structured legal facts from the sentence. For example, "Zhang San intentionally injured Li Si" can be extracted as the triple: (Zhang San, intentionally injured, Li Si).

[0028] Relationship Extraction Algorithm: This part combines a convolutional neural network and a conditional random field model to automatically identify legal relationships in the text. For example, in complex compound sentences, the relationships between various legal entities are identified through a deep learning model.

[0029] After the input legal text is processed by the text preprocessing unit, it enters the semantic analysis unit for syntactic parsing and semantic role labeling, and finally the structured legal fact triples are extracted through the relationship extraction unit. The whole process can be illustrated by the following example: Input: "Zhang San intentionally injured Li Si, causing Li Si minor injuries, and the court sentenced Zhang San to be guilty."

[0030] After being processed by the text preprocessing unit: Word segmentation: ["Zhang San", "intentionally", "injured", "Li Si", "causing", "minor injuries", "court", "sentenced", "guilty"]; Part-of-speech tagging: [Zhang San / person's name, deliberately / verb, injure / verb, Li Si / person's name, minor injury / noun, court / noun, sentence / verb, guilty / adjective]; Named entity recognition: The entities recognized are: Zhang San (the subject), Li Si (the victim), and the court (the institution); After semantic analysis unit parsing: Syntactic tree analysis: It is determined that the main sentence structure is "Zhang San deliberately injured Li Si", and the result is a subject-verb-object relationship.

[0031] Semantic role labeling: "Zhang San" is the agent of the action, "deliberately injured" is the action, and "Li Si" is the victim.

[0032] The relationship extraction unit outputs structured triples: (Zhang San, deliberately injured, Li Si), (the court, sentenced, Zhang San guilty).

[0033] The core task of the legal knowledge graph construction module is to construct a hierarchical and inferable knowledge graph based on the legal facts extracted by the input processing module to support applications such as legal retrieval, reasoning analysis, and case recommendation. This module uses technologies such as entity linking, relationship fusion, and graph structure optimization to convert the structured data in legal texts into a computable legal knowledge network.

[0034] In the legal field, the construction of a knowledge graph needs to fully consider the hierarchical relationship of legal provisions, the causal chain of case facts, and the precedent effect of judicial judgments. This module not only integrates existing laws and regulations but also can perform automated structured processing on newly added legal texts through machine learning models to achieve continuous expansion and update of legal knowledge.

[0035] This module consists of multiple sub-units, including the legal entity recognition and alignment unit, the legal relationship extraction unit, and the knowledge fusion and storage unit. The technical solutions of each unit are as follows: 4. Legal entity recognition and alignment unit: Based on the legal facts extracted by the input processing module, this unit further optimizes entity recognition and performs cross-context entity alignment to ensure the consistency of the same legal concept in different cases or articles.

[0036] Entity standardization: Normalize the entities in legal texts. For example, "Contract Law of the People's Republic of China" can be aligned with "Contract Law" to ensure the uniqueness of nodes in the knowledge graph.

[0037] Entity disambiguation: For ambiguous legal terms, use the context-based BERT embedding matching algorithm and combine the legal knowledge base for disambiguation. For example, "guarantee" may refer to "mortgage guarantee" or "surety guarantee" in different legal contexts, and this module can accurately distinguish through context information.

[0038] Suppose the input legal text contains the following entities: "Contract"; "Contract Law"; "Contract Law of the People's Republic of China"; "Contract Volume of the Civil Code"; Use an entity alignment model based on embedding vectors to calculate the similarity between different entities: ; Among them, and are the embedding vectors of the entities. If the calculated similarity exceeds a set threshold (such as 0.9), it is determined to be the same entity and merged when stored in the knowledge graph.

[0039] 5. Legal relationship extraction unit: The task of this unit is to identify the legal relationships between entities and establish the edge structure of the knowledge graph according to the logic of legal provisions.

[0040] Rule-driven legal relationship mapping: Establish relationships between common legal entities through predefined legal relationship rules. For example, there is a "constraint" relationship between "Contract" and "Liability for Breach of Contract", which can be expressed as: ; Relationship reasoning based on graph neural networks: For legal relationships that are not clearly stipulated but have implicit logic, use the GNN model to learn the association patterns in legal provisions and infer potential legal relationships. For example, learn the potential connection between "Fraudulent Act" and "Invalidity of Contract" from multiple case data.

[0041] 6. Knowledge fusion and storage unit: After completing the extraction of legal relationships, this unit fuses all legal entities and their relationships and stores them as queryable knowledge graph data.

[0042] For different legal relationships, the following relationship types are defined in the knowledge graph: Constraint relationship (such as "Contract" constrains "Liability for Breach of Contract"); Causal relationship (such as "Fraud" causes "Invalidity of Contract"); Citation relationship (such as "Civil Code" cites "Contract Law"); For complex legal relationships, use a GNN model based on the attention mechanism to learn the hidden associations in case data. Set the adjacency matrix A of the legal knowledge graph, the entity embedding matrix, and H is updated through the GNN propagation formula: ; Among them: Denote the entity embedding matrix of the th layer; are the parameters trained by the model; is the activation function (such as ReLU). This calculation method can learn the potential patterns between legal relationships and improve the relationship reasoning ability.

[0043] Knowledge fusion strategy: Adopt the RDF (triple) storage model to represent legal knowledge in the structure of (entity - relationship - entity), for example: ; Graph database storage: Select a database (such as Neo4j) to store the legal knowledge graph to support efficient legal query and reasoning.

[0044] Input: "According to the Contract Law, a sales contract shall clearly define the liability for breach of contract. If the breaching party fails to perform its obligations, it shall bear the liability for compensation." Knowledge graph construction process: Legal entity recognition: Identify "Contract Law", "Sales contract", "Liability for breach of contract", "Liability for compensation"; Legal relationship extraction: (Contract Law, stipulates, Sales contract)(Contract Law, stipulates, Sales contract) (Sales contract, specifies, Liability for breach of contract)(Sales contract, specifies, Liability for breach of contract) (Liability for breach of contract, leads to, Liability for compensation)(Liability for breach of contract, leads to, Liability for compensation) Knowledge storage: Store the triples into the knowledge graph for subsequent query and reasoning; Through the above technical solutions, the legal knowledge graph can transform legal texts into a structured and inferable knowledge network, support legal retrieval and intelligent reasoning, and automatically expand the knowledge boundary through machine learning technology. This module is not only applicable to a single legal system but also can adapt to the legal knowledge expression in different legal domains through cross - context entity alignment technology.

[0045] The main goal of the automatic theorem - proving module is to conduct an automated analysis of the provability of legal propositions based on the structured legal data of the legal knowledge graph construction module, combined with logical reasoning and mathematical deduction methods. This module can identify the premises and conclusions in the input legal propositions and establish a proof chain through a formal logical system to achieve applications such as legal reasoning, case adjudication assistance, and legal education.

[0046] In practical applications, legal reasoning is different from traditional mathematical theorem proving, and its logical system often includes characteristics such as uncertain reasoning, inductive reasoning, and case-based reasoning. Therefore, this module adopts a proof method that combines first-order logic (FOL), modal logic, and inductive logic, and introduces an automated theorem prover (ATP), such as Prover9, Coq, or Lean, to ensure the completeness and verifiability of the reasoning process.

[0047] This module consists of multiple subunits, including a legal proposition formalization unit, an inference rule generation unit, and a proof engine invocation unit. The technical solutions of each unit are as follows: 7. Legal proposition formalization unit: The main function of this unit is to convert the input legal proposition into a formal logical expression for subsequent processing by the proof engine.

[0048] Fact extraction based on the knowledge graph: Extract relevant legal provisions and case facts from the legal knowledge graph. For example, for the proposition "What are the conditions for the establishment of a sales contract?", the following can be found in the knowledge graph: ; First-order logic conversion: Convert the legal proposition into the form of first-order predicate logic (FOL). For example: , , contract( , ) ⇒ (offer( ) ∧ acceptance( ))), where contract( , ) represents and 's sales contract relationship, offer( ) represents makes an offer, and acceptance( ) represents accepts the offer.

[0049] This unit automatically generates inference rules applicable to automated theorem proving according to the legal logical system.

[0050] Rule mining based on inductive reasoning: Extract high-frequency legal reasoning patterns from historical case data through machine learning techniques. For example, in contract law cases, the frequently occurring inference rule: (contract valid ∧ unperformed) ⇒ liability for breach of contract (contract valid ∧ unperformed) ⇒ liability for breach of contract Modal logic extension: Judgments of "possibility" or "necessity" are often involved in legal reasoning, so modal logic is used to extend the rule set.

[0051] 8. Proof engine invocation unit This unit calculates formal propositions and inference rules, invokes an automated theorem prover (ATP) tool for proof, and returns the proof path and derivation result.

[0052] Based on the method of resolution theorem proving, using ATP tools such as Prover9, it derives whether a proposition holds through resolution deduction. For example, given the following premises: , sales contract( ) ⇒ (offer( ) ∧ acceptance( )) If the proposition: sales contract(contract 1) then the proof system can automatically derive: offer(contract 1) ∧ acceptance(contract 1); Furthermore, it confirms whether the conditions for the contract to be established are met.

[0053] Proof method based on satisfiability (SAT) solving: For decision-making problems in legal issues (such as whether a certain law applies to a specific case), it can be converted into a SAT problem for solution. For example, for the proposition: ; Through the SAT solver, it can determine whether the proposition is satisfiable, thereby determining its logical feasibility.

[0054] Input case: "A and B signed a sales contract, B did not pay the goods, can A request B to pay liquidated damages?" Legal reasoning process: Extract relevant legal rules: (contract established ∧ ¬performance) ⇒ liability for breach of contract(contract established ∧ ¬performance) ⇒ liability for breach of contract Set the premise: contract established(contract AB), ¬payment(B) contract established(contract AB), ¬payment(B) Inference result: liability for breach of contract(B) liability for breach of contract(B); Finally, it is derived that B should bear the liability for breach of contract, and further reasoning is carried out in combination with the knowledge graph on whether the liquidated damages clause applies.

[0055] The main goal of the reasoning explanation module is to make the reasoning process of the automated theorem proving module transparent, providing interpretable, traceable, and verifiable reasoning results. This module can not only display the reasoning path, but also visually present the legal rules, logical deduction process, and reasoning basis involved in the reasoning process, so as to enhance the understandability of legal reasoning and provide intuitive support for legal retrieval, case determination, and legal learning.

[0056] In the process of legal reasoning, the automated theorem proving module usually conducts reasoning and deduction using first-order logic (FOL), modal logic, and inductive reasoning. However, the complexity of legal reasoning makes it difficult for users to directly understand the process of logical derivation. Therefore, this module realizes the transparency of the reasoning process by combining technologies such as natural language generation (NLG), graph structure visualization, and construction of traceable reasoning chains, making legal reasoning not only computable but also interpretable.

[0057] This module relies on the legal knowledge graph and the derivation results of automated theorem proving to conduct a structured analysis of the reasoning process, and combines the hierarchical structure of legal rules to provide logical verification and case support for the reasoning process to ensure the traceability of the reasoning conclusion.

[0058] This module consists of multiple subunits, including a reasoning path analysis unit, a reasoning chain construction unit, and an interpretability enhancement unit. The technical solutions of each unit are as follows: Reasoning path analysis unit The main function of this unit is to analyze the reasoning process of the automated theorem proving module and extract key reasoning steps, derivation formulas, and legal bases so that users can clearly understand the reasoning process.

[0059] Path analysis based on resolution reasoning: In the process of automated theorem proving, if resolution theorem proving is adopted, this unit can analyze its derivation tree. For example, for a contract reasoning problem: ; If the input premise is: Contract established (Contract A), ¬Performed (Contract A) Contract established (Contract A), ¬Performed (Contract A) This unit can analyze the derivation path and output: Step 1: Confirm the contract is established Contract established (Contract A) Contract established (Contract A) Step 2: The contract is not performed ¬Performed (Contract A) ¬Performed (Contract A) Step 3: Derive the conclusion Liability for breach of contract (Contract A) Liability for breach of contract (Contract A) Modal logic path analysis: When legal reasoning involves judgments of "possibility" or "necessity", this unit can analyze modal logic derivation. This unit can analyze: Possible situation: The contract may be invalid (legal uncertainty); Necessary derivation: If the contract is invalid, liability must be waived; This unit is responsible for constructing a reasoning chain to make the legal reasoning process traceable, and combines the legal knowledge graph to provide legal provisions or case precedents to support the reasoning.

[0060] Inference Chain Storage Based on Graph Database: Use a graph database (such as Neo4j) to store the inference process, enabling the inference chain to have query and visualization capabilities.

[0061] Case Support and Inference Chain Supplement: When the inference involves case law, this unit can automatically retrieve similar cases. For example: Rule Inference: Failure to perform a contract leads to liability for breach of contract Case Supplement: Based on a certain case (Case Number: 2023 Contract Judgment No. 001), the court determined that the buyer's failure to pay the purchase price shall bear liability for breach of contract.

[0062] 8. Interpretability Enhancement Unit The core task of this unit is to convert the inference path into a readable natural language description and display it in combination with a visualization tool to enhance the interpretability of the inference.

[0063] Inference Explanation Based on Natural Language Generation (NLG): This unit can automatically generate natural language descriptions to make the inference path easy to understand. For example: Input logical expression: .

[0064] Output explanation: According to the provisions of the Contract Law, if a contract has been established but one party fails to perform, that party shall bear liability for breach of contract. Therefore, in this case, since Contract A has been established and the contractual obligations have not been performed, it can be deduced that liability for breach of contract is established.

[0065] Inference Visualization: This unit supports converting the inference path into a tree structure or a knowledge graph.

[0066] Input case: A and B signed a sales contract, and B did not pay the purchase price. Can A require B to pay liquidated damages? Extract the inference chain: (1) The sales contract has been established; (2) B did not pay the purchase price, constituting a failure to perform the contract; (3) According to the provisions of the Contract Law, failure to perform contractual obligations shall bear liability for breach of contract; (4) Liability for breach of contract may include payment of liquidated damages.

[0067] Visualize the inference path: (Sales Contract)--[Established]--(Valid Contract); (Valid Contract)--[Not Performed]--(Liability for Breach of Contract); (Liability for Breach of Contract)--[Legal Basis]--(Contract Law); Automatically generate an explanation: The sales contract between Party A and Party B has been established. According to the provisions of the Contract Law, if a party to a contract fails to perform its obligations, it shall bear the liability for breach of contract. In this case, Party B has not paid the purchase price, which constitutes a failure to perform the contract. Therefore, Party A has the right to require Party B to pay liquidated damages.

[0068] The reasoning and explanation module can structurally analyze the reasoning process of automatic theorem proving and, in combination with the legal knowledge graph, provide a traceable reasoning chain, making legal reasoning not only computable but also interpretable. This module realizes the transparency of legal reasoning and improves the comprehensibility and verifiability of legal reasoning in practical applications through techniques such as resolution reasoning analysis, reasoning chain storage, natural language generation, and reasoning visualization.

[0069] The main goal of the cross-jurisdictional conflict resolution module is to provide automated analysis and solutions for conflicts between different legal systems or legal norms based on the legal knowledge graph, automatic theorem proving, and reasoning and explanation module. This module can identify legal application conflicts between different jurisdictions (such as different countries, regions, legal systems) and, in combination with legal reasoning, hierarchical analysis of norms, and logical consistency verification, derive the optimal legal application strategy.

[0070] In practical applications, cross-jurisdictional legal conflicts usually involve international law, conflict of laws, treaty application rules, and the comparison of laws in different countries or regions. Due to differences in the binding force of legal norms in different jurisdictions, conflict resolution requires not only logical reasoning but also consideration of legal effect, priority, and scope of application. Therefore, this module combines hierarchical analysis of legal rules, resolution logical reasoning, SAT-based satisfiability analysis, and legal application judgment of modal logic to realize automated analysis of cross-jurisdictional conflicts and generation of solutions.

[0071] The core functions of this module include: identification of jurisdictional conflicts, legal application reasoning, recommendation of conflict solutions, and, in combination with the knowledge graph and case analysis, ensuring the rigor and interpretability of the legal reasoning process.

[0072] This module consists of multiple subunits, including a jurisdictional conflict identification unit, a legal application reasoning unit, and a conflict solution recommendation unit. The technical solutions of each unit are as follows: Jurisdictional conflict identification unit The main function of this unit is to identify legal application conflicts between different jurisdictions and classify the types of conflicts for subsequent reasoning and processing.

[0073] Conflict mining based on the knowledge graph: Using the legal knowledge graph, extract possible conflict rules from legal texts in different jurisdictions. For example, assume that the law of a certain country stipulates: ; While the law of another jurisdiction stipulates: ; This unit can automatically detect this contradiction and mark it as "conflict in law application".

[0074] Analysis of legal domain constraints based on modal logic: Legal application rules usually contain characteristics such as "possibly applicable" or "must be applicable", and need to be analyzed in combination with modal logic. For example: ; This formula means that "if the law of legal domain A must be applicable, then the law of legal domain B cannot be applicable". This unit can parse such constraints and form conflict recognition rules.

[0075] 9. Legal application reasoning unit This unit is used to conduct logical reasoning on the identified legal domain conflicts, and combined with the hierarchical structure of legal rules, deduce more applicable legal rules.

[0076] Applicability reasoning based on the legal hierarchical structure: Legal rules usually have different levels, such as international treaties, domestic laws, local regulations, case laws, etc. This unit uses logical reasoning to judge the application priority. For example, if there are: International treaty ⇒ Contract application rule 1; Domestic law ⇒ Contract application rule 2; Then this unit determines through reasoning and analysis whether the international treaty has the right of prior application over domestic law.

[0077] Legal consistency verification based on SAT solving: If the rules of two legal domains are contradictory, they can be converted into a satisfiability problem (SAT) for analysis. For example: ; If the SAT solver determines that this formula is unsatisfiable, it means that there are irreconcilable conflicts in the rules of the two legal domains, and this unit needs to further deduce the application priority.

[0078] Applicability analysis combined with case law: In the actual process of legal application, some conflicts can be explained through case law. For example: Case law in legal domain A supports the enforcement of contracts Case law in legal domain B supports liquidated damages in lieu of performance If the case law in legal domain A has higher efficacy in the international judicial system, then the rules of legal domain A shall be preferentially applicable 10. Conflict solution recommendation unit This unit is used to provide specific legal application suggestions based on the reasoning results, and combined with the legal knowledge graph, recommend relevant cases or legal interpretations.

[0079] Legal application recommendation based on resolution reasoning: This unit uses resolution reasoning to deduce the optimal legal application plan. For example: ; If the rules of legal jurisdiction A are not applicable, the rules of legal jurisdiction B will be automatically recommended for application.

[0080] Generation of legal application strategies for multiple legal jurisdictions: In some cases of legal conflicts, this unit can generate more compatible legal application strategies. For example: Combine the case laws of legal jurisdiction A and legal jurisdiction B to form a compromise application rule; Apply the principles of conflict of laws (such as the doctrine of the most significant relationship) to make rulings on legal application; Combine historical case data to recommend legal rules with a higher probability of application; Case recommendation and legal interpretation: This unit can provide legal basis for conflict resolution based on the legal knowledge graph. For example: If a case involves a cross-border contract dispute, this unit can retrieve similar cases, such as relevant case laws of the United Nations Commission on International Trade Law; Automatically generate explanatory text. For example: Since the contract involves cross-border transactions, according to the doctrine of the most significant relationship, the law of the place where the contract was signed should be applied. In addition, referring to a reference case (case number: 2024 Commercial International Judgment No. 005), the court ruled that the United Nations Convention on Contracts for the International Sale of Goods should be applied.

[0081] Input case: Company A (legal jurisdiction A) and Company B (legal jurisdiction B) signed a contract. Company A claims specific performance of the contract, but Company B claims only payment of liquidated damages according to its national law. How to resolve this conflict? Cross-jurisdictional conflict resolution process: Identification of legal jurisdiction conflicts: Legal jurisdiction A: Contract performance ⇒⇒ Specific performance; Legal jurisdiction B: Contract performance ⇒⇒ Substitute performance with liquidated damages; Marking of rule conflicts: Applicable rules are contradictory; Legal application reasoning: Apply international conventions (such as CISG): Contract performance should comply with the provisions of international conventions; Applicability analysis: International convention ⇒⇒ Specific performance takes precedence; Resolution reasoning: CISG ⇒ Specific performance applicable to CISG ⇒ Specific performance applicable; Recommended solutions: Solution 1: Apply international convention, and Company A can claim specific performance of the contract Solution 2: If Company B provides sufficient reasons (such as excessive performance costs), the liquidated damages rule can be applied in a compromise manner Recommended case law: A certain international commercial arbitration case supports the precedence of specific performance; The core objective of the adaptive optimization module is to provide continuous and dynamic optimization strategies based on the aforementioned legal reasoning, cross-jurisdictional conflict resolution, and reasoning explanation modules, so as to improve the reasoning efficiency and accuracy of the system in different environments. Through real-time analysis and feedback mechanisms, this module adaptively adjusts various parameters and rules in the reasoning process based on user input, system reasoning results, and changes in the external environment, thereby enhancing the intelligence and adaptability of the overall legal reasoning system.

[0082] In practical applications, legal reasoning tasks often face uncertainties and complexities, involving a large number of legal rules, case precedents, and their interactions. To address this challenge, this module combines machine learning optimization algorithms, dynamic rule adjustment mechanisms, and feedback learning mechanisms, and can automatically adjust and optimize in different legal jurisdictions, case types, and legal environments to ensure the efficiency and accuracy of reasoning results.

[0083] Specifically, the adaptive optimization module can dynamically optimize the reasoning chain, adjust the rule hierarchy based on the reasoning processes and results of the aforementioned modules, and continuously improve the reasoning strategy through learning from historical data, thereby providing legal advice and decision-making support that better conforms to the actual situation.

[0084] The adaptive optimization module mainly consists of three sub-units: a dynamic rule adjustment unit, an optimization feedback learning unit, and a real-time reasoning optimization unit. The specific technical implementations of each unit are as follows: 11. Dynamic rule adjustment unit: The main function of this unit is to dynamically adjust the priorities and application strategies of legal rules based on the real-time requirements of reasoning tasks to optimize the reasoning effect.

[0085] Rule adjustment based on reasoning task analysis: By analyzing the case information input by the user and the reasoning task, the applicable order of rules is automatically adjusted. For example, for a specific case, certain case laws or international treaties may need to be applied preferentially, while for other cases, local regulations may be more important. This unit can make real-time adjustments based on the relevance and priority of the rules. For example, when facing contract performance issues, relevant provisions of the Contract Law may be considered preferentially, while in international commercial cases, international conventions may need to be applied first.

[0086] Rule optimization based on reasoning complexity: In a multi-level and multi-rule reasoning process, the application order and complexity of rules will affect the reasoning efficiency. This unit can analyze the complexity in the current reasoning process in real time and optimize the use of rules according to the analysis results. For example, when the reasoning path is complex, the system may automatically select more concise reasoning rules to avoid complex logical operations, improving the system response speed and computing efficiency.

[0087] 12. Optimization feedback learning unit The core function of this unit is to optimize the execution of future reasoning tasks through feedback and learning from historical reasoning results. By analyzing historical data, this unit can achieve continuous self - adjustment during the reasoning process, thereby improving the accuracy and efficiency of the overall reasoning system.

[0088] Feedback learning based on machine learning: By analyzing successful and failed cases in the historical reasoning process through machine learning algorithms, optimization patterns in the reasoning process are extracted and applied to future reasoning tasks. For example, when the system analyzes that a certain type of case always involves conflicts of certain legal rules, the system will learn how to automatically detect and resolve such conflicts, improving the accuracy and adaptability of the reasoning results.

[0089] Optimization of reasoning results based on historical case law: Through deep learning of historical case law, the system can optimize its case - based reasoning process. For example, if certain cases often relied on specific legal provisions or case laws in the past, the system can, based on this historical trend, preferentially apply these case laws or legal provisions, thereby improving the accuracy and decision - making speed of the reasoning results.

[0090] 13. Real - time Reasoning Optimization Unit This unit is mainly responsible for dynamically adjusting reasoning strategies during the reasoning process to improve the accuracy and efficiency of reasoning. Specifically, this unit can combine real - time input data and environmental conditions to instantaneously optimize the reasoning process.

[0091] Dynamic optimization based on input data: When the user inputs new case information or modifies existing case parameters, this unit can instantaneously adjust relevant rules and strategies in the reasoning process. For example, when a certain piece of data in the case changes, the system can adjust the reasoning path through adaptive optimization to ensure the validity and accuracy of the reasoning results.

[0092] Reasoning adjustment based on external environment changes: In different legal jurisdictions or different legal environments, the application of laws and reasoning strategies may vary. This unit can sense external environment changes in real - time and adjust reasoning strategies according to new legal provisions. For example, in some legal jurisdictions, new case laws may cause changes in previous reasoning paths, and the system can automatically update reasoning rules and paths according to this change.

[0093] Input case: Company A and Company B signed an international trade contract. Company B failed to make payment on time, and Company A demanded payment of liquidated damages. The system needs to determine whether Company A can demand payment of liquidated damages and recommend the optimal reasoning path.

[0094] Adaptive optimization processing flow: Input analysis: After the system receives the case information, it first analyzes relevant legal provisions such as the Contract Law and international conventions to identify applicable rules.

[0095] Based on the historical case library, analyze whether liquidated damages are usually applicable and whether there are conflicts of legal domains in similar cases.

[0096] Dynamic rule adjustment: The system automatically adjusts the rule priorities and gives priority to applying relevant provisions in international conventions.

[0097] If there are special liquidated damages regulations in the legal domain where Company B is located, the system will adjust according to the laws of that legal domain to determine whether to apply the regulations of the local legal domain.

[0098] Optimization feedback learning: By analyzing historical data, the system finds that in similar cases, liquidated damages are usually applicable when the contract is not performed and there are no special circumstances. Therefore, the system automatically increases the application probability of the liquidated damages rule.

[0099] Real-time reasoning optimization: If during the processing, Company B claims that its payment obligation has been affected by force majeure factors, the system will update the reasoning path in real time, add the force majeure rule, and make corresponding reasoning adjustments.

[0100] Finally, it is inferred that if Company B fails to provide valid force majeure certificates, Company A has the right to claim liquidated damages.

[0101] The core goal of the VR teaching terminal is to build a visual and interactive legal teaching environment through virtual reality (VR) technology, immersive interaction systems, and intelligent teaching engines. Based on the aforementioned legal reasoning system, adaptive optimization module, and cross-legal domain conflict resolution solutions, this module provides multi-dimensional and multi-level legal education support, enabling users to intuitively understand legal rules, case analysis, and complex reasoning processes in a virtual environment.

[0102] This terminal is not only applicable to the teaching scenarios of law schools but also can be used in various professional fields such as judicial training, lawyer mock trials, and cross-legal domain legal application drills. By combining VR visualization technology, voice interaction, intelligent Q&A, and case demonstrations, this module can enhance the intuitiveness, interactivity, and depth of legal learning.

[0103] In terms of specific implementation, this terminal can provide core functions such as legal knowledge visualization, VR mock court, multi-role interaction experience, and real-time reasoning feedback, and combine intelligent knowledge graphs, semantic parsing, and speech recognition technologies to enable users to conduct multi-scenario legal learning and practical training in an immersive environment.

[0104] The VR teaching terminal consists of sub-units such as the legal knowledge visualization unit, immersive interaction unit, case simulation demonstration unit, intelligent teaching engine unit, etc. The technical solutions of each unit are as follows: 14. Legal Knowledge Visualization Unit The core goal of this unit is to structure and visualize legal knowledge, presenting the complex legal system in an intuitive way to enhance users' understanding and memory.

[0105] Legal visualization based on knowledge graph: This unit uses a knowledge graph to construct a structured representation of the legal system, including legal provisions, case laws, legal domain relationships, reasoning chains, etc. Users can intuitively browse the legal system in a VR environment. For example: Show the logical relationship between legal provisions in the form of a three-dimensional network diagram; Show the priority of legal application in a hierarchical tree structure; Combined with a timeline, show the evolution process of case law; Visualization of legal application paths based on modal logic: The legal application paths of different cases may vary. This unit combines modal logic to deduce the application paths and presents them in a visual way. For example: If a user inputs an international commercial contract dispute case, the system can display the legal application situations of different legal domains through a visual path, such as: ; This formula can be transformed into an animation display of the path applicable to legal domain A to help users understand the application logic.

[0106] 15. Immersive Interaction Unit The core function of this unit is to achieve multi-mode human-computer interaction through VR devices, enabling users to have an immersive experience in a virtual legal environment.

[0107] Legal Q&A based on speech recognition and natural language processing: This unit combines speech recognition with a legal knowledge base, allowing users to ask questions by voice. The system can parse them in real time and give legal answers. For example: The user asks: "There is no liquidated damages stipulated in the contract. Can I claim compensation from the other party?" The system can automatically search for relevant legal provisions and display the liability for breach of contract rules in the Contract Law through voice and VR text, while marking the applicable legal provisions in a three-dimensional interface.

[0108] Multi-role interaction experience: Users can choose different roles (such as lawyer, judge, party) to participate in legal simulations. For example: Users can act as lawyers and conduct debates in a virtual courtroom; The system can automatically generate the other party's debate strategies and adjust the case analysis process in real time.

[0109] 16. Case Simulation and Demonstration Unit The core objective of this unit is to use VR technology to conduct dynamic demonstrations of legal cases, enabling users to intuitively experience the development process of cases in a virtual environment and engage in interactive learning.

[0110] Case Demonstration Based on Timeline: This unit combines historical judicial precedents to simulate the entire process of case occurrence, trial, and adjudication. For example, when simulating a contract dispute case: The system can sequentially display the signing of the contract, breach of contract behavior, litigation process, and adjudication results, and allow users to select different legal application strategies at different stages and observe their impacts.

[0111] Dynamic Presentation of Legal Reasoning Process: This unit can dynamically demonstrate complex legal reasoning processes. For example: In the issue of legal application across legal domains, the system can display the conflicts in legal applications of different legal domains and demonstrate the resolution process in 3D animation. For example: Step 1: Display the conflict rules of legal domain A and legal domain B; Step 2: Dynamically calculate the applicable priority; Step 3: Demonstrate the applicable path of the final adjudication.

[0112] 17. Intelligent Teaching Engine Unit The core function of this unit is to combine artificial intelligence to achieve intelligent legal teaching guidance, ensuring that users at different levels can obtain appropriate teaching content.

[0113] Dynamic Teaching Strategies Based on Knowledge Level: This unit can dynamically adjust teaching content according to users' learning progress and knowledge mastery. For example: Beginner Mode: Provide basic legal concepts and accompany them with case animations; Advanced Mode: Provide the legal reasoning process and require users to participate in reasoning; Professional Mode: Users need to conduct legal analysis by themselves, and the system will conduct intelligent scoring; Personalized Reasoning Task Training: This unit combines an adaptive optimization module to automatically generate personalized reasoning training tasks. For example: If a user has a good grasp of "Contract Law" but is weak in "International Commercial Law", the system will give priority to recommending cross-border contract dispute cases for training and dynamically adjust the reasoning difficulty.

[0114] Input Case: The user hopes to learn international contract law and participate in legal reasoning exercises for contract breach cases.

[0115] VR Teaching Process: Case Visualization: The system displays the structure of international contracts in a VR environment and marks key terms.

[0116] Combined with a knowledge graph, it demonstrates the core principles of contract law and applicable cases.

[0117] Interactive learning: The user selects the "contract breach" topic through voice or gesture operations.

[0118] The system shows different scenarios of breach and allows the user to select applicable legal provisions.

[0119] Case reasoning exercise: The user enters a virtual courtroom, plays the role of a lawyer, and defends the client.

[0120] The system generates the debate content of the opposing lawyer, and the user needs to put forward rebuttals based on contract law.

[0121] Combined with an AI reasoning engine, the system evaluates the user's legal reasoning ability in real time and provides improvement suggestions.

[0122] Working principle: During the operation of the system, the case information input by the user is first processed by the legal reasoning system. Based on multi-source data such as legal provisions, case law, and international conventions, this system generates legal reasoning paths through techniques such as modal logic reasoning, semantic parsing, and legal knowledge graph construction, and conducts rationality judgment and legal application analysis in combination with the user input. To address legal conflict issues in different legal domains, the present invention also designs a cross-legal-domain conflict resolution module, which realizes intelligent reasoning and application under different legal systems through methods such as rule priority judgment and legal application path reconstruction.

[0123] During the legal reasoning process, to improve the intelligence and adaptability of the system, the present invention introduces an adaptive optimization module. This module continuously self-optimizes the legal reasoning process through three major mechanisms: dynamic rule adjustment, optimization feedback learning, and real-time reasoning optimization. The dynamic rule adjustment unit automatically adjusts the priority and application strategy of rules according to the characteristics and reasoning complexity of the current case; the optimization feedback learning unit analyzes historical reasoning results through machine learning algorithms and continuously improves the reasoning strategy; the real-time reasoning optimization unit dynamically adjusts the reasoning path and rule application in combination with the user input and changes in the external environment to ensure the accuracy and efficiency of the reasoning results.

[0124] At the same time, the VR teaching terminal of the present invention realizes the visualization and interactive learning of legal knowledge and legal reasoning process through the collaborative work of legal knowledge visualization, immersive interaction, case simulation demonstration and intelligent teaching engine. The legal knowledge visualization unit presents legal provisions, case law and legal relations in a structured three-dimensional environment, so that users can intuitively understand the legal system and its applicable logic. The immersive interaction unit combines voice recognition and natural language processing technology to provide real-time legal Q&A and multi-role simulation experience, thereby enhancing the interactivity and fun of learning. The case simulation demonstration unit dynamically displays the trial and reasoning process of legal cases through VR technology, allowing users to conduct full-process legal practice exercises in a virtual environment. The intelligent teaching engine unit makes personalized adjustments based on the user's learning data, provides adaptive legal learning plans for users at different levels, and guides users' learning and progress through intelligent scoring and feedback mechanisms.

[0125] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence-based immersive legal education system, characterized in that, Including: An input processing module configured to perform multi-level semantic analysis on the legal text input by the user and generate a structured legal fact representation; A legal knowledge graph construction module connected to the input processing module for receiving structured legal facts and constructing and updating a legal knowledge graph; An automatic theorem proving module connected to the legal knowledge graph construction module for performing multi-hop reasoning on the legal knowledge graph based on probabilistic reasoning; A reasoning explanation module connected to the automatic theorem proving module for visually displaying the reasoning process using explainable artificial intelligence technology; A cross-jurisdictional conflict resolution module connected to the legal knowledge graph construction module for detecting and resolving cross-jurisdictional legal provision conflicts; An adaptive optimization module connected to the cross-jurisdictional conflict resolution module and the automatic theorem proving module respectively for dynamically optimizing the reasoning parameters and the knowledge graph structure based on user feedback and system performance metrics; A VR teaching terminal connected to the reasoning explanation module for generating an immersive teaching environment including a three-dimensional virtual court scene, a dynamic case evolution process, and an interactive teaching interface.

2. The immersive legal education system based on artificial intelligence according to claim 1, wherein The input processing module includes: A text preprocessing unit configured to perform word segmentation, part-of-speech tagging, and named entity recognition on the input text; A semantic analysis unit connected to the text preprocessing unit and configured to perform syntactic analysis and semantic role labeling using a BiLSTM model based on an attention mechanism; A relation extraction unit connected to the semantic analysis unit and configured to extract legal fact triples based on a predefined legal relation template.

3. The immersive legal education system based on artificial intelligence according to claim 1, characterized in that, The legal knowledge graph construction module includes: An entity relation encoding layer configured to perform feature encoding on legal provision nodes using a graph convolutional network; A cross-jurisdictional association layer connected to the entity relation encoding layer and configured to establish a semantic mapping relationship between legal provisions in different jurisdictions through a contrastive learning algorithm; A dynamic update layer connected to the cross-jurisdictional association layer and configured to update the knowledge graph in real time to reflect the revision of legal provisions.

4. An artificial intelligence-based immersive legal education system according to claim 3, characterized in that, The cross-jurisdictional association layer includes: A jurisdiction feature extraction unit configured to extract the legal provision features of a specific jurisdiction; A semantic alignment unit connected to the jurisdiction feature extraction unit and configured to minimize the embedding distance of similar legal concepts; A bridging edge generation unit connected to the semantic alignment unit and configured to establish cross-jurisdictional association edges.

5. The immersive legal education system based on artificial intelligence according to claim 1, characterized in that, The automatic theorem proving module includes: A predicate logic conversion unit configured to convert structured legal facts into predicate logic expressions including legal subjects, actions, and objects; A multi-hop reasoning unit connected to the predicate logic conversion unit and configured to perform multi-hop reasoning based on an attention mechanism on the legal knowledge graph; An evidence credibility evaluation unit connected to the multi-hop reasoning unit and configured to calculate the confidence of each reasoning path based on a Bayesian reasoning framework.

6. The immersive legal education system based on artificial intelligence according to claim 1, characterized in that, The reasoning explanation module includes: A reasoning path visualization unit configured to generate a tree-like topology graph including a legal provision citation chain; An influence degree analysis unit connected to the reasoning path visualization unit and configured to quantify the contribution degree of each legal element to the reasoning result using SHAP values.

7. An immersive legal education system based on artificial intelligence according to claim 1, characterized in that, The cross-jurisdictional conflict resolution module includes: A conflict detection unit configured to identify legal provisions with applicable conflicts; A conflict resolution unit connected to the conflict detection unit and configured to calculate the applicability score of provisions based on the semantic similarity of provisions, the citation frequency of cases, and the timeliness of legal provisions.

8. An artificial intelligence-based immersive legal education system according to claim 1, characterized in that, The adaptive optimization module includes: A reinforcement learning agent configured to dynamically adjust the evidence weight parameter, the fact similarity weight, and the conflict resolution coefficient according to user feedback using the PPO algorithm; A knowledge graph updater connected to the reinforcement learning agent and configured to update the knowledge graph structure in real time.

9. The legal immersive education system based on artificial intelligence according to claim 1, characterized in that, The VR teaching terminal includes: A gesture recognition unit configured to capture the operation instructions of the user on the virtual legal document; A dynamic case evolution unit connected to the gesture recognition unit and configured to change the debate process of the virtual court in real time according to the reasoning path selected by the user; A teaching effect evaluation unit connected to the dynamic case evolution unit and configured to calculate the attention distribution of the user on the key legal elements through eye tracking data.

Images