Civil affair judgment result prediction method and system based on thinking chain

Abstract

The invention discloses a civil affair judgment result prediction method and system based on a thinking chain, and belongs to the technical field of natural language processing. The invention aims to utilize the thinking chain technology, fully utilize the case facts and the information between the original toll name and defendant deferences, take the right request as the starting point, take entity legal specification as the basic means to form essential analysis, and disassemble the legal logic of civil trial to obtain the legal logic of the civil trial. A complete civil case judgment process is formed through a series of thinking with a logical relationship, challenges caused by diversity and complexity of civil cases are effectively dealt with, and it is ensured that civil judgment result prediction reaches a higher level in accuracy and logic leakproofness.

Description

Technical Field

[0001] The present invention relates to the field of natural language processing technology, and more specifically, to a method and system for predicting civil judgment results based on thought chains. Background Art

[0002] The core goal of applying artificial intelligence and big data technologies in the judicial field is to comprehensively improve the efficiency and transparency of judicial work with the help of advanced intelligent technologies.

[0003] In the legal field, giving professional legal judgments is an important part of legal practice. Civil cases often involve a variety of legal relationships, such as contractual relationships, tort relationships, property rights relationships, etc. Each relationship has its own unique legal provisions and judgment rules, making the automated prediction of legal judgment results more complicated. At present, there are two main methods for predicting civil judgment results: one is based on mathematics, statistical algorithms, machine learning and other methods. These methods are based on legal knowledge and case characteristics to build suitable legal models. They have high requirements for data accuracy and completeness and are often only applicable to legal cases in specific scenarios. When faced with new and complex cases, these methods are difficult to have legal logical reasoning mechanisms and have the disadvantage of insufficient generalization ability. The second is a prediction method based on historical similar cases. This method predicts the possible judgment results of the current case by analyzing the results of similar cases in the past. It relies on the collection, collation and analysis of a large number of historical cases to discover the similarities and regularities between cases. However, the judgment of civil cases cannot rely entirely on empirical data, and its legitimacy should not be based solely on the regularity of past cases. Summary of the invention

[0004] In order to overcome the above problems, the present invention provides a method for predicting civil judgment results based on thought chain. The present invention aims to utilize thought chain technology, make full use of the information between the facts of the case and the plaintiff's claims and the defendant's defenses, take the rights request as the starting point, and analyze the constituent elements of substantive legal norms as the basic means to dismantle the legal logic of civil trials, form a complete civil case trial process through a series of logically related thinking, effectively respond to the challenges brought by the diversity and complexity of civil cases, and ensure that the prediction of civil judgment results reaches a higher level in accuracy and logical rigor.

[0005] To achieve the above objectives, the technical solution of the present invention includes the following contents.

[0006] A method for predicting civil judgment results based on thought chain, the method comprising:

[0007] Obtaining input information and extracting case elements from the input information; wherein the input information includes: claims, defenses and case facts;

[0008] Search the knowledge base based on the input information and case elements to obtain historical cases and relevant laws and regulations similar to the input information;

[0009] Combine the claim and case elements to build the first chain of thought for the purpose of fixing the claim and specifying the claim, call the big model, and obtain the claim;

[0010] Combining the claims, defenses, case elements, rights requests, historical cases similar to the input information, and relevant laws and regulations, a second thinking chain is constructed to screen the basic norms of rights requests and the basic norms of defense rights, and the large model is called to obtain the basic norms of rights requests, the basic norms of defense rights, and the defense claims and reasons;

[0011] Combining the basic norms of rights claims and the basic norms of defense rights, constructing a third thinking chain for the purpose of decomposing the constituent elements of the basic norms, calling the big model, and obtaining the constituent elements of the basic norms of rights claims and the constituent elements of the basic norms of defense rights;

[0012] Combining the claims and reasons of the defense, the basic normative elements of the claims and the basic normative elements of the defense, we build the fourth thinking chain for the purpose of generating the controversial focus of the case, and call the big model to obtain the controversial focus of the case.

[0013] In combination with the case elements, historical cases similar to the input information, rights claims, claims and reasons of the defense, basic norms for rights claims, basic norms for rights of defense, constituent elements of basic norms for rights claims and constituent elements of basic norms for rights of defense, and the controversial focus of the case, a fifth thinking chain is constructed for the purpose of classifying the elements and generating the judgment results, and the big model is called to obtain the civil judgment prediction results of the input information; wherein, the civil judgment prediction results include: the content of the judgment reasoning and the content of the judgment conclusion.

[0014] Furthermore, extracting case elements from the input information includes:

[0015] Extract litigation requests based on the claims;

[0016] Extract defense opinions based on the arguments;

[0017] Extract events based on claims, defenses, and case facts;

[0018] Extract key information based on claims, defenses and case facts.

[0019] Furthermore, the key information includes: civil acts, right holders, liability holders, types of payments, amounts, quantities, performance holders, performance contents and performance methods.

[0020] Furthermore, the knowledge base includes: a civil case recognition knowledge base, which consists of a first plaintext part and a first vector part. The first plaintext part includes civil historical cases and the abstracts, case features, and legal provision contents of these civil historical cases. The first vector part is generated based on the semantic information of the civil historical cases;

[0021] Retrieve the civil case recognition knowledge base according to the input information and case elements to obtain historical cases similar to the input information, including:

[0022] Based on the input information, generate a vector representation of the input information. Based on the input information and its vector representation, and in combination with the first plaintext part and the first vector part, obtain a first matching result;

[0023] Based on the case elements, generate a vector representation of the case elements. Based on the case elements and their vector representation, and in combination with the first plaintext part and the first vector part, obtain a second matching result;

[0024] Based on the first matching result and the second matching result, obtain historical cases similar to the input information.

[0025] Furthermore, the knowledge base includes: a civil laws and regulations knowledge base, which consists of a second plaintext part and a second vector part. The second plaintext part includes: civil legal provisions, judicial interpretations, and the characteristic labels and themes of these civil legal provisions and judicial interpretations. The second vector part is generated based on the semantic information of the civil legal provisions and the judicial interpretations;

[0026] Retrieve the civil laws and regulations knowledge base according to the input information and case elements to obtain legal regulations related to the input information, including:

[0027] Based on the input information, generate a vector representation of the input information. Based on the input information and its vector representation, and in combination with the second plaintext part and the second vector part, obtain a third matching result;

[0028] Based on the case elements, generate a vector representation of the case elements. Based on the case elements and their vector representation, and in combination with the second plaintext part and the second vector part, obtain a fourth matching result;

[0029] Based on the third matching result and the fourth matching result, obtain legal regulations related to the input information.

[0030] Furthermore, the first thinking chain, the second thinking chain, the third thinking chain, the fourth thinking chain, and the fifth thinking chain also include: detailed reference contents helpful for achieving the corresponding task objectives.

[0031] Furthermore, the training process of the large model includes:

[0032] Construct training datasets based on data from general and judicial fields;

[0033] The LoRA fine-tuning method is used on the training data to fine-tune the large model.

[0034] A civil judgment result prediction system based on thought chain, the system comprising:

[0035] An information acquisition module is used to acquire input information and extract case elements from the input information; wherein the input information includes: claims, defenses and case facts;

[0036] The knowledge retrieval module is used to search the knowledge base based on the input information and case elements to obtain historical cases and relevant laws and regulations similar to the input information;

[0037] The rights request generation module is used to combine the claims and case elements to construct the first thinking chain for the purpose of fixing the claim and specifying the rights request, calling the big model, and obtaining the rights request;

[0038] The basic norms generation module is used to combine the claims, defenses, case elements, rights requests, historical cases similar to the input information, and relevant laws and regulations to build a second thinking chain for the purpose of screening the basic norms of rights requests and the basic norms of defense rights, and call the big model to obtain the basic norms of rights requests, the basic norms of defense rights, and the defense claims and reasons;

[0039] The basic normative constituent elements generation module is used to combine the basic norms of the right claim and the basic norms of the right of defense, build a third thinking chain for the purpose of decomposing the basic normative constituent elements, call the big model, and obtain the basic normative constituent elements of the right claim and the basic normative constituent elements of the right of defense;

[0040] The module for generating the focus of disputes in a case is used to combine the claims, the claims and reasons of the defense, the constituent elements of the basic rules of the claims and the constituent elements of the basic rules of the defense, to construct the fourth thinking chain for the purpose of generating the focus of disputes in the case, and to call the big model to obtain the focus of disputes in the case;

[0041] The judgment result prediction module is used to combine the case elements, historical cases similar to the input information, rights claims, claims and reasons of the defense, basic norms of rights claims, basic norms of right of defense, constituent elements of basic norms of rights claims and constituent elements of basic norms of right of defense, and the controversial focus of the case, to construct the fifth thinking chain for the purpose of classifying elements and generating judgment results, call the big model, and obtain the civil judgment prediction results of the input information; wherein, the civil judgment prediction results include: the content of the judgment reasoning and the content of the judgment conclusion.

[0042] An electronic device, comprising: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements any of the above-mentioned methods for predicting civil judgment results based on thought chains.

[0043] A computer-readable storage medium, characterized in that computer program instructions are stored on the computer-readable storage medium, and when the computer program instructions are executed by a processor, any of the above-mentioned civil judgment result prediction methods based on thought chain is implemented.

[0044] Compared with the prior art, the present invention has at least the following beneficial effects.

[0045] The present invention uses the thinking chain technology to disassemble and reorganize complex legal logic into a series of interrelated thinking steps, making the entire prediction process more in line with the essential characteristics of legal reasoning and enhancing the logic and rationality of the prediction results. And compared with the traditional method that relies on specific scenarios or a large number of historical cases, the present invention starts from the case itself and analyzes and predicts by constructing a reasonable logical chain, which can better handle newly emerging complex case types and show stronger generalization ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is the overall flow chart of the present invention.

[0047] Figure 2 It is a schematic diagram of the technical route of the present invention. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present invention will be fully described below in conjunction with the accompanying drawings in the implementation of the present invention.

[0049] The overall process of the present invention is shown in Figure 1 , technical route see Figure 2 , including the following steps:

[0050] S1. Input the claim, defense and case facts;

[0051] S2. Call the big model to extract case elements from the input information.

[0052] Case elements include litigation request extraction, defense opinion extraction, event extraction, and key information extraction. Among them, litigation request extraction and defense opinion extraction include the extraction of specific requirements or claims raised by the plaintiff and the defendant in the case; event extraction includes the extraction of events in three parts: claim, defense, and case facts; key information extraction includes civil acts, right holders, liability holders, types of payment, amounts, quantities, performers, performance contents, and performance methods, etc.

[0053] S3. Based on the input claims and defenses, case facts, and case elements, a hybrid retrieval method of text similarity matching based on BM25 (Best Matching) and semantic similarity matching based on SIMCES (Similarity Contrastive Estimation) is used to search the civil case cognition knowledge base and the civil laws and regulations knowledge base to obtain historical cases and relevant laws and regulations similar to the input information.

[0054] The civil case cognition knowledge base includes a plain text part and a vector storage part. Specifically, the plain text part includes civil historical cases (the case causes are civil cases issued by the Supreme People's Court, but do not include special procedures, non-litigation procedures, and undisclosed case causes, and the cases are civil cases that occurred after the promulgation of the Civil Code of the People's Republic of China) and the part after knowledge enhancement processing; these civil historical cases are subjected to knowledge enhancement processing based on a large model, including summary generation, case feature extraction, and legal content recognition, and indexing with civil historical cases; SIMCES-based vector representation technology is used to construct a vector library to store semantic information to form the final civil case cognition knowledge base;

[0055] The civil laws and regulations knowledge base includes a plain text part and a vector storage part. Specifically, the plain text part includes civil law provisions, judicial interpretations, and the part after knowledge enhancement processing of civil law provisions and judicial interpretations. The civil law provisions include the Civil Code of the People's Republic of China, the Arbitration Law of the People's Republic of China, the Civil Procedure Law of the People's Republic of China, the Partnership Enterprise Law, the Company Law, the Labor Law, the Labor Contract Law, the Consumer Rights Protection Law, etc. The judicial interpretations include the Interpretation of the Supreme People's Court on the Application of the Tort Liability Code of the Civil Code of the People's Republic of China (I), the Interpretation of the Supreme People's Court on Several Issues Concerning the Application of the General Principles of the Contract Code of the Civil Code of the People's Republic of China, and the Provisions of the Supreme People's Court on Several Issues Concerning the Application of Laws in the Trial of Private Lending Cases. The overall structural analysis of the legal and regulatory texts is carried out to split the hierarchical entries. These legal provisions are subjected to knowledge enhancement processing based on large models, including feature label extraction and theme extraction, and indexing with legal and regulatory provisions. The vector storage part is constructed using SIMCES-based vector representation technology to store semantic information to form the final civil laws and regulations knowledge base.

[0056] S4. Combine the claims and case elements to construct a thinking chain aimed at fixing the claim and concretizing the claim, call the big model, and finally obtain the claim.

[0057] The reference template of the thinking chain is as follows:

[0058] Input section:

[0059] “Role: Civil judge;

[0060] Task: [Specific task description and specific requirements for fixing rights claims, including but not limited to analysis, reasoning, decision-making, etc.]

[0061] Known information:

[0062] (1) Allegation:

[0063] (2) Case elements:

[0064] Reference Information:

[0065] (1) Related reference information 1: [Detailed reference content that helps to complete the task objectives];

[0066] (2) Related reference information 2: [Detailed reference content that helps to complete the task objectives];

[0067] Analytical reasoning: [This step requires careful consideration of known information, combined with reference information, and thinking in accordance with the legal trial logic of civil cases, and ultimately reasoning to obtain the rights claim;]

[0068] Final Fixed Rights Request:

[0069] Summary and confirmation: [Finally fix the plaintiff's claim and clarify the specific content and composition of the plaintiff's claim.]"

[0070] Output section:

[0071] "Rights Requested:\n{}"

[0072] Wherein, {} represents the input claim information and the case elements obtained in step S2. The reference information includes legal logic knowledge related to fixing the plaintiff's claim and clarifying the plaintiff's specific claim, aiming to provide support for the fixation of the claim.

[0073] In one embodiment, a specific example of a thought chain is as follows:

[0074] Input section:

[0075] “Role: Civil judge;

[0076] Task: Fully analyze the known information such as the claim and case elements, fully learn from and refer to the reference information to fix the plaintiff's litigation request, clarify the reference information of the plaintiff's specific litigation request, etc., and finally fix the plaintiff's litigation request and clarify the plaintiff's rights request (the specific content and composition of the litigation request). Note that the rights request cannot have defects such as contradictions and ambiguity;

[0077] Known information:

[0078] (1) Allegation:

[0079] (2) Case elements:

[0080] Reference information:

[0081] (1) Fix the reference information of the plaintiff's litigation request: the right to request others to act or not act, including the right to claim for debt, the right to return, the right to damages, the right to compensation and recourse, the right to reimbursement of expenses, the right to inaction, the right to personal rights, and other rights; clarify the right to form, which is the right to establish, change or eliminate legal relations, including the right to request changes due to major misunderstandings or obvious unfairness, the right to change due to fraud or coercion, the right of set-off under contract law, the right to revoke under contract law, the right to revoke company resolutions, the right of first refusal under company law, the right of first refusal under contract law, the right of first refusal under property law, and the right of co-ownership. Limited purchase rights, contract cancellation rights, termination rights, etc.; clarify the adversarial and complementary relationships of rights, and pay special attention to the fact that different types of lawsuits and different requests may form adversarial or complementary relationships; clarify the competition of rights, and pay attention to whether the same fact meets the requirements of different legal norms at the same time, including competition of laws (the same event meets the constituent elements of more than two legal norms at the same time, but only one of the legal norms can be applied according to the rules of legal application), selective competition (when the same event meets two or more basic rights norms at the same time, the right holder can only choose one of the rights between the right to request and the right to form), aggregation of claims, and competition of claims;

[0082] (2) Reference information to clarify the plaintiff's specific litigation request: ① When the plaintiff requests confirmation of the validity of a civil act, it should be clear whether it is valid, invalid, ineffective, or revoked; ② When the plaintiff requests confirmation of the ownership of rights, the right holder, nature, and content should be clearly stated; ③ When the plaintiff requests payment of money or physical objects, the payment liability subject, type, amount, and quantity should be clearly stated; ④ When the plaintiff requests the performance of a certain act, the performance subject, content, and method of performance should be clearly stated;

[0083] Analytical reasoning: This step requires careful consideration of the plaintiff's opinion statement (i.e., the claim, known information item 1), combined with the specific circumstances of the case facts (known information item 2) and reference information, and the following aspects of thinking in a logical order, and finally reasoning to obtain the rights claim;

[0084] (1) Determine the type of claim: Based on the aforementioned preparations and the part of the claim that fixes the plaintiff's litigation request in the reference information (reference information item 1), determine which type of claim the plaintiff's rights claim belongs to, or whether it involves the issue of formative rights. If there is a conflict of rights, the conflict of rights must also be clarified. Finally, determine the type of claim.

[0085] (2) Specify the claim: Based on the above preparations and the part of the reference information that specifies the plaintiff's specific claim (reference information item 2), all the above analysis results are finally converted into a specific statement of the claim, and finally the claim is obtained. For example: if you request to confirm the validity / invalidity of a certain civil act, you need to clarify the status of the act; if you seek to determine the ownership of a specific property right, you need to specify the object of the right ownership and related details; when requiring the other party to pay money or other forms of compensation, you must list the required amount and the list of items in detail; if you hope to enforce a specific action, you must clearly state who the executor is, what action should be taken, and how it will be implemented.

[0086] Final Fixed Rights Request:

[0087] Summary and confirmation: Based on the above analysis, the plaintiff’s claim is finally confirmed, and the specific content and composition of the plaintiff’s claim are clarified.

[0088] Output section:

[0089] "Rights Requested:\n{}"

[0090] S5. Combine the claims, defenses, case elements, the rights request obtained in step S4, historical cases similar to the input information obtained in step S3, and relevant laws and regulations to construct a thinking chain for the purpose of screening the basic norms for rights requests and the basic norms for defenses (rights), call the big model, and obtain the basic norms for rights requests, the basic norms for defenses, and the defense claims and reasons.

[0091] The reference template of the thinking chain is as follows:

[0092] Input part:

[0093] “Role: Civil judge;

[0094] Task: [Specific task description and specific requirements with the goal of clarifying the defense claims and reasons, screening the basic norms of rights claims and the basic norms of defense (rights), including but not limited to analysis, reasoning, decision-making and other aspects;]

[0095] Known information:

[0096] (1) Allegation:

[0097] (2) Argument:

[0098] (3) Case elements:

[0099] (4) Claim:

[0100] Reference Information:

[0101] (1) Similar historical cases:

[0102] (2) Relevant laws and regulations:

[0103] (3) Related reference information 1: [Detailed reference content that helps to complete the task objectives];

[0104] (4) Related reference information 2: [Detailed reference content that helps to complete the task objectives];

[0105] Analytical reasoning: [This step requires careful consideration of known information, combined with reference information, and thinking in accordance with the legal trial logic of civil cases. Finally, the defense claims and reasons are obtained through reasoning, and the basic norms of rights claims and defense (rights) are screened out;]

[0106] Finally, the basic specifications were selected:

[0107] Summary and confirmation: [Finally, the defense claims and reasons are obtained through reasoning, and the basic norms of rights claims and defense (rights) are obtained through screening.]”

[0108] Output section:

[0109] "Claim base specification:\n{}\n;

[0110] Defense claims and reasons:\n{}\n;

[0111] Basic norms for defense (right):\n{}”

[0112] Where {} represents the input claim, defense information, case elements obtained in step S2, and the rights request obtained in step S4. The reference information includes historical cases and relevant laws and regulations similar to the input information obtained in step S3, and legal logic knowledge related to clarifying the basic norms of rights request and clarifying the basic norms of defense (rights), aiming to provide support for obtaining the basic norms of rights request, basic norms of defense rights, and defense claims and reasons.

[0113] In one embodiment, a specific example of a thought chain is as follows:

[0114] Input section:

[0115] “Role: Civil judge;

[0116] Task: Fully analyze the known information such as the claim, defense, case elements, and rights requests, fully learn from and refer to similar historical cases, similar legal and regulatory provisions, reference information that clarifies the basic norms of rights requests, reference information that clarifies the basic norms of defense (rights), etc., and finally clarify the claims and reasons of the parties' defenses, and clarify the substantive law norms corresponding to the defenses, that is, the basic norms of defense (rights); focus on the rights requests, and finally list the laws and regulations that support the plaintiff's litigation requests, that is, the basic norms of rights requests;

[0117] Known information:

[0118] (1) Allegation:

[0119] (2) Argument:

[0120] (3) Case elements:

[0121] (4) Claim:

[0122] Reference information:

[0123] (1) Similar historical cases:

[0124] (2) Relevant laws and regulations:

[0125] (3) Clarify the reference information of the basic norms of the rights request: refers to the legal norms that support the plaintiff's litigation request. The purpose is to list in detail the laws and regulations that support the plaintiff's litigation request based on the rights request. For example: if the rights request involves requesting the confirmation of property rights (including ownership, usufruct rights, security rights and other property rights), the corresponding basic norm is Article 234 of the Civil Code: If there is a dispute over the ownership and content of the property right, the interested party may request confirmation of the right. If the request is to confirm the mortgage right, the provisions of Article 402 and Article 403 of the Civil Code should also be included.

[0126] (4) Clarify reference information on the basic norms of defense (right): The right of defense is the right to oppose the claim, and is the defendant's right to refuse to pay when the plaintiff requests the plaintiff to exercise the right. Check whether the defendant has raised a defense or exercised the right of defense. If the defendant has raised a defense or exercised the right of defense, the right of defense should be clarified. According to the time and scope of effect, the right of defense includes permanent defense (permanently preventing the claim from taking effect, that is, the claimant's lawsuit for payment will be considered groundless and rejected), temporary defense (temporarily preventing the court from enforcing the claim, that is, the claimant's lawsuit for payment will be temporarily rejected as unreasonable, and can only temporarily exclude the other party's claim within a certain period of time), and restrictive defense (cannot prevent the court from enforcing the claim, but can lead to a payment judgment that gradually accepts that the claimant should pay, and limitedly supports the claimant's claim); on the basis of clarifying the right of defense, further clarify the basic norms of defense (right). The basic norms of defense (right) are the legal norms on which the defendant raises the defense (right). First, make sure whether the defendant's defense claims or reasons are clear. Second, identify whether the defendant has raised a defense (right) in his defense or has the intention to defend. Third, based on the right of defense, list in detail the laws and regulations to which the right of defense refers.

[0127] Analytical reasoning: This step requires careful consideration of the specific circumstances of the claims, defenses, case elements, and rights requests (known information), and combined with similar historical cases and relevant laws and regulations (reference information items 1 and 2), the following aspects should be considered in a logical order, and ultimately the content and composition of the parties' defense, the basic norms for rights requests, and the basic norms for defense rights should be deduced.

[0128] (1) Reference to similar historical cases: Based on the claims, defenses, case elements, rights claims, etc. (known information), refer to the legal provisions mentioned in similar historical cases (reference information item 1), especially the provisions directly related to the core issues of the case, and retain the most relevant legal provisions.

[0129] (2) Reference to relevant laws and regulations: Based on the content of the claim, defense, rights request, and right basis (known information), refer to relevant laws and regulations (reference information item 2), and retain the most relevant laws and regulations.

[0130] (3) Determine the basic norms for the claim: Based on the above analysis results, according to the part of the reference information that clarifies the basic norms for the claim (reference information item 3) and the legal provisions retained in (1) and (2), and combined with the applicability of the specific circumstances of the case, finally list the laws and regulations that support the plaintiff's claim, that is, the basic norms for the claim. For example, if the case involves a loan contract dispute, the legal provisions related to borrowing and guarantee may be involved to judge the case;

[0131] (4) Determine the basic norms of defense (rights): Based on the above analysis results, analyze the defendant's reply based on the part of the reference information that clarifies the basic norms of defense (rights) (reference information item 4) and the legal provisions retained in (1) and (2), and determine whether there are valid reasons for defense or rights of defense. If so, further explore whether the defense complies with the law and find out the specific legal provisions that support its position. The key to this step is to identify all possible conditions that can effectively counter the plaintiff's claims and understand how these conditions interact to form an effective defense strategy, and finally clarify the claims and reasons of the parties' defenses, and clarify the substantive law norms corresponding to the defense, that is, the basic norms of defense (rights).

[0132] Finally, the basic specifications were selected:

[0133] Summary and confirmation: Based on the above analysis, the claims and reasons of the parties’ defense are finally clarified, and the substantive law norms corresponding to the defense, that is, the basic norms of defense (rights); focusing on the rights request, the laws and regulations supporting the plaintiff’s litigation request are finally listed, that is, the basic norms of the rights request. "

[0134] Output section:

[0135] "Claim basis:\n{}\n;

[0136] Defense claims and reasons:\n{}\n;

[0137] Basic norms for defense (right):\n{}”

[0138] S6. Combined with the basic norms obtained in step S5, construct a thinking chain aimed at decomposing the constituent elements of the basic norms, call the big model, and obtain the constituent elements of the basic norms of rights claims and the basic norms of defense (rights).

[0139] The reference template of the thinking chain is as follows:

[0140] Input section:

[0141] “Role: Civil judge;

[0142] Task: [Specific task description and specific requirements aimed at decomposing the basic norms of rights claims and basic norms of defenses, including but not limited to analysis, reasoning, decision-making, etc.]

[0143] Known information:

[0144] (1) Basic specifications for claiming rights:

[0145] (2) Basic norms of defense (right):

[0146] Reference information for decomposing the constituent elements of the basic specification:

[0147] [Detailed reference content that helps to complete the task objectives];

[0148] Analytical reasoning: [This step requires careful consideration of known information, combined with reference information, and thinking in accordance with the legal trial logic of civil cases, and ultimately reasoning to obtain the decomposition of the constituent elements of the basic norms of the right claim and the basic norms of the defense (right);]

[0149] Final output basic specification components:

[0150] Summary and confirmation: [The final reasoning leads to the decomposition of the constituent elements of the basic norms of the right claim and the basic norms of the defense (right).]”

[0151] Output section:

[0152] "Basic normative elements of the claim:\n{}\n;

[0153] Basic normative elements of defense (right):\n{}”

[0154] Wherein, {} represents the basic norms of the claim and the basic norms of the defense (right) obtained in step 5. The reference information includes legal logic knowledge related to the decomposition of the constituent elements of the basic norms, and is intended to provide support for obtaining the constituent elements of the basic norms of the claim and the basic norms of the defense (right).

[0155] In one embodiment, a specific example of a thought chain is as follows:

[0156] Input part:

[0157] “Role: Civil judge;

[0158] Task: Fully analyze the basic norms of rights claims and the basic norms of defense (rights), fully learn from and refer to the reference information of the decomposition of the constituent elements of the basic norms, analyze the constituent elements contained in all relevant legal provisions, and finally obtain the constituent elements of the basic norms of rights claims and the basic norms of defense (rights);

[0159] Known information:

[0160] (1) Basic specifications for claiming rights:

[0161] (2) Basic norms of defense (right):

[0162] Reference information for decomposing the constituent elements of the basic specification:

[0163] The basic norms supporting the plaintiff's litigation request and the defendant's defense (right) reasons are analyzed according to the logic of legal norms. The legal provisions are divided into complete legal provisions and incomplete legal provisions based on whether they can be used as an independent basis for the right to claim. Complete legal provisions are legal provisions that can serve as an independent basis for the right to claim. They are composed of two parts: constituent elements and legal effects. For example, if one party fails to perform its contractual obligations or performs its contractual obligations inconsistent with the agreement; the legal effect is that it should bear the liability for breach of contract such as continuing to perform, taking remedial measures or compensating for losses; incomplete legal provisions refer to legal provisions that cannot be directly used as the basis for the right to claim and need to be linked to other legal provisions. Assumptions in legal provisions usually have basic constituent elements, which are constituent elements. There are two basic structural forms of constituent elements:

[0164] Structure 1: M1+M2+M3......=R

[0165] In the above formula, M1, M2, M3, etc. represent different constituent elements. R represents the legal effect.

[0166] For example, Article 34 of the Property Law states: "If a person has no right to occupy immovable or movable property, the right holder may request the return of the original property."

[0167] M1: The occupant has no rights.

[0168] M2: The possessor has taken possession of the immovable or movable property.

[0169] Legal effect (R): The right holder can request the return of the original property.

[0170] The analysis of the elements of the basic norms is the process of analyzing the elements and implicit elements. Only when all the elements are examined clearly can the legal provisions be accurately applied to make fair judgments.

[0171] Structure 2: M1, M2 or M3......=R

[0172] For example, in Article 107 of the Contract Law, the first requirement is: "One party fails to perform its contractual obligations"; the second requirement is: "or the performance of the contractual obligations is not in compliance with the agreement." These two requirements are optional conditions.

[0173] Analytical reasoning: Think about the following aspects in a logical order, and finally deduce the constituent elements of the basic norms of rights claims and the constituent elements of the basic norms of defenses (rights).

[0174] (1) Clarify the constituent elements of the basic norms for the claim: Based on the known basic norms for the claim (known information item 1), fully refer to the reference information for decomposing the constituent elements of the basic norms (reference information item 1) to identify whether the norm is a complete legal provision or an incomplete legal provision. According to the principle of decomposing the constituent elements of the basic norms, if it is a complete legal provision, directly extract the constituent elements M1, M2, ... and the legal effect R from it; if it is an incomplete legal provision, it means that as long as one of the conditions is met, the legal effect can be produced. Finally, the constituent elements of the basic norms for the claim are obtained.

[0175] (2) Clarify the constituent elements of the basic norms of the defense (right): Similarly, for the basic norms on which the defendant’s defense is based (known information item 2), similar decomposition of constituent elements is also required. Finally, the constituent elements of the basic norms of the defense (right) are obtained.

[0176] Final output basic specification components:

[0177] Summary and confirmation: Based on the above analysis, we finally get the constituent elements of the basic norms of rights claims and the basic norms of defenses (rights). "

[0178] Output section:

[0179] "Basic elements of the claim:\n{}\n;

[0180] Basic normative elements of defense (right):\n{}”

[0181] S7. Combine the rights claim obtained in step S4, the claims and reasons for the defense obtained in step S5, and the basic regulatory elements of the rights claim and the basic regulatory elements of the defense (right) obtained in step S6, and construct a thinking chain for the purpose of generating the focus of the dispute of the case, call the big model, and obtain the focus of the dispute of the case.

[0182] The reference template of the thinking chain is as follows:

[0183] Input part:

[0184] “Role: Civil judge;

[0185] Task: [Specific task description and specific requirements for generating litigation issues, including but not limited to analysis, reasoning, decision-making, etc.]

[0186] Known information:

[0187] (1) Claim:

[0188] (2) Defense claims and reasons:

[0189] (3) Constituent elements of the basic norms of the claim:

[0190] (4) Constitutive elements of the basic norm of the right of defense: {}

[0191] Reference information for sorting out the key points of litigation disputes:

[0192]

Detailed reference content helpful for achieving the task objectives.

[0193] Analysis and reasoning:

In this link, it is necessary to carefully consider the known information, combine the reference information, and think according to the legal trial logic of civil cases, and finally reason out the key points of the case disputes.

[0194] Finally output the key points of the case disputes:

[0195] Summary and confirmation:

Finally reason out the key points of the case disputes.

[0196] Output part:

[0197] “Key points of the case disputes: \n{}”

[0198] Among them, {} represents the claim right obtained in step S4, the claim and reasons of the defense obtained in step S5, and the constitutive elements of the basic norm of the claim right and the constitutive elements of the basic norm of the right of defense obtained in step S6. The reference information includes the legal logic knowledge related to sorting out the key points of litigation disputes, aiming to provide support for obtaining the key points of the case disputes.

[0199] In one embodiment, a specific example of the chain of thought is as follows:

[0200] Input part:

[0201] “Role: Civil judge;

[0202] Task: Sufficiently analyze the case elements, claim rights, defense claims and reasons, fully draw on and refer to the reference information for sorting out the key points of litigation disputes, and through judicial acts such as comparing, analyzing, summarizing and clarifying the claim rights and defense claims, clarify the conflicts existing in the litigation claims of the plaintiff and defendant and the basic norms of their claim rights, and on this basis, finally determine the key points of the case disputes, noting that the key points of disputes should be relevant to the claims;

[0203] Known information:

[0204] (1) Claim right: {}

[0205] (2) Defense claim and reasons: {}

[0206] (3) Constitutive elements of the basic norm of the claim right: {}

[0207] (4) Constitutive elements of the basic norm of the right of defense: {}

[0208] Reference information for organizing the focus of litigation disputes:

[0209] According to the claims and defenses of the plaintiff and the defendant and their basic norms, the disputed points of the case are summarized and the trial is carried out around the disputed points. Disputes refer to specific matters that the parties have disputes, including matters of both law and fact. The arrangement of disputed points should be based on the constituent elements of basic norms. The arrangement of disputed points can effectively limit the scope of the trial and speed up the trial efficiency. The arrangement of disputed points follows the principle of "one party's claim - whether the other party admits (self-admits) - evidence and cross-examination when not admitting - determination based on the rules for the allocation of the burden of proof".

[0210] Analytical reasoning: Think in the following logical order, and finally deduce the focus of the dispute, the reasoning of the judgment, and the conclusion of the judgment. Based on the claims and defenses of both parties (known information items 1 and 2) and the constituent elements of the basic norms of their claims and the constituent elements of the basic norms of their defenses (known information items 3 and 4), fully refer to and organize the reference information of the litigation points, find out the differences in the positions of both parties, and finally summarize the focus of the dispute of the case;

[0211] The focus of the dispute in the final output case:

[0212] Summary and confirmation: Based on the above analysis, we finally get the controversial focus of the case. "

[0213] Output section:

[0214] "The controversial focus of the case:\n{}"

[0215] S8. Combine the case elements, similar historical cases obtained in step S3, the rights claim obtained in step S4, the claims and reasons of the defense obtained in step S5, the basic norms of the rights claim and the basic norms of the defense (right), the constituent elements of the basic norms of the rights claim and the constituent elements of the basic norms of the defense (right) obtained in step S6, and the focus of the dispute of the case obtained in step S7, and construct a thinking chain for the purpose of classifying the elements and generating the judgment results, call the big model, and obtain the reasoning and civil judgment results of the civil judgment.

[0216] The reference template of the thinking chain is as follows:

[0217] Input part:

[0218] “Role: Civil judge;

[0219] Task: [Specific task description and specific requirements for classifying elements and generating adjudication results, including but not limited to analysis, reasoning, decision-making, etc.]

[0220] Known information:

[0221] (1) Case elements:

[0222] (2) Claim:

[0223] (3) Basic norms for claiming rights:

[0224] (4) Defense claims and reasons:

[0225] (5) Basic norms of defense (right):

[0226] (6) Constituent elements of the basic norms of the claim:

[0227] (7) Constituent elements of the basic norms of defense (right):

[0228] (8) The focus of the dispute in the case:

[0229] Reference information:

[0230] (1) Similar historical cases:

[0231] (2) Related reference information 1: [Detailed reference content that helps to complete the task objectives];

[0232] (3) Related reference information 2: [Detailed reference content that helps to complete the task objectives];

[0233] Analytical reasoning to generate judgment reasoning and judgment conclusions: [This step requires careful consideration of known information, combined with reference information, and thinking in accordance with the legal trial logic of civil cases, and ultimately reasoning to obtain civil judgment reasoning and civil judgment results;]

[0234] The final output is the judgement reasoning content and judgement conclusion:

[0235] Summary and confirmation: [The final reasoning resulted in the civil judgment reasoning and civil judgment results.]”

[0236] Output section:

[0237] "Content of the referee's reasoning:\n{}\n;

[0238] Content of the judgement conclusion:\n{}”

[0239] Among them, {} represents the case elements, the claim obtained in step S4, the basic norms of the claim obtained in step S5, the claims and reasons of the reply and the basic norms of the defense (right), the constituent elements of the basic norms of the claim and the constituent elements of the basic norms of the defense (right) obtained in step S6, the controversial focus of the case obtained in step S7, and the similar historical cases obtained in step S3. The reference information includes historical cases similar to the input information obtained in step S3, legal logic knowledge related to the classification of elements, civil judgments and logical syllogisms, and is intended to provide support for generating judgment reasoning and judgment conclusions.

[0240] In one embodiment, a specific example of a thought chain is as follows:

[0241] Input section:

[0242] “Role: Civil judge;

[0243] Task: Fully analyze the case elements, rights claims, defense claims and reasons, dispute focus, basic norms, basic norm constituent elements, etc., fully learn from and refer to similar historical cases and reference information for the classification of elements and making judgments, compare and classify the events in the case elements with the constituent elements of the basic norms of the rights claims and the constituent elements of the basic norms of the defense (right), and finally form the content of the judgment reasoning and obtain the judgment result through the classification process of the reference logical syllogism;

[0244] Known information:

[0245] (1) Case elements:

[0246] (2) Claim:

[0247] (3) Basic norms for claiming rights:

[0248] (4) Defense claims and reasons:

[0249] (5) Basic norms of defense (right):

[0250] (6) Constituent elements of the basic norms of the claim:

[0251] (7) Constituent elements of the basic norms of defense (right):

[0252] (8) The focus of the dispute in the case:

[0253] Reference information:

[0254] (1) Similar historical cases:

[0255] (2) Classification of elements and making a judgment: Classify the facts of the case into the constituent elements of legal norms, that is, legal facts and constituent elements facts. There are three requirements for completing the classification: First, the facts of the case must be facts that have been legally recognized, that is, the facts of the case are consistent with the requirements of the basic elements of the constituent elements; second, the facts that are consistent with the constituent elements must not only meet the individual legal constituent elements, but also all the elements; third, the opponent's defense is not established, in other words, the defense (right) must be invalid before the legal basis norms can be applied to support the plaintiff's litigation request. As long as one element cannot be classified, the legal provision should not be used to support the right request (defense request). The right of defense also needs to be legally classified, and the classification method is exactly the same as the classification method of the claim right, and all conditions must be met. Compare the constituent elements of the legal norms with the recognized facts. There are several elements in the legal provision, and it is necessary to determine whether each element is established and must be recognized. If each element fact can be recognized, it can be legally classified, that is, the legal provision can be applied to make a judgment. The structural form of the classification of elements: the constituent elements are: M1, M2, M3...; the case facts are: A1, A2, A3...; then the case facts A1, A2, A3...can be classified according to the law only if the defense is not established and all the legal constituent elements of M1, M2, M3... are met.

[0256] (3) Logical syllogism: A syllogism consists of three parts: major premise, minor premise and conclusion. The major premise is the legal provision or legal principle, that is, the legal article or legal interpretation on which the judgment is based. The minor premise is the specific facts of the case. The conclusion is the result derived from the major premise and minor premise, that is, the judgment result.

[0257] Generate analytical reasoning for the judge's reasoning and conclusion: Think about the following aspects in a logical order, and ultimately deduce the focus of the dispute, the judge's reasoning, and the judge's conclusion.

[0258] First, fully refer to the reference information for the classification of elements and making judgments (reference information item 2), and correspond each event in the case elements and defense claims and reasons (known information items 1 and 4) one by one to the constituent elements of the basic norms of the right claim and the constituent elements of the basic norms of the defense (right) (known information items 6 and 7), and conduct comparative analysis. When comparing, first, ensure that each basic norm constituent element can be met, and second, ensure that the defense is established; second, combine the case elements, basic norms, and the focus of the case (known information items 1, 6, 7, and 8), and make a judgment reasoning on whether to apply or not apply the legal provisions based on the classification results; finally, strictly follow the method of logical syllogism (reference information item 3) to finally obtain the judgment conclusion of the case. The writing style and ideas of the judgment reasoning and judgment conclusion can refer to the judgment results of similar historical cases (reference information item 1). Finally, the judgment reasoning and judgment conclusion are obtained.

[0259] The final output is the judgement reasoning content and judgement conclusion:

[0260] Summary and confirmation: Based on the above analysis, we finally get the content of the judgment reasoning and the judgment conclusion. "

[0261] Output section:

[0262] "Content of the referee's reasoning:\n{}\n;

[0263] Content of the judgement conclusion:\n{}”

[0264] Among them, the big model mentioned in steps S2, S3, S4, S5, S6, S7 and S8 is a fine-tuned big model in the legal field. The implementation method includes: combining general field and judicial field data, using hash deduplication, semantic coherence detection and other technologies, constructing case summary, dispute focus generation, reasoning generation and other adjudication task knowledge and legal theory data for 800,000 judicial training corpora, based on the 14B basic model (Qwen1.5-14B), using the LoRA fine-tuning method, to form a fine-tuned big model in the legal field.

[0265] It should be clear that the specific implementation and drawings of the present invention disclosed above are intended to help understand the content of the present invention and implement it accordingly. For those of ordinary skill in the art, it can be understood that various substitutions, changes and modifications are within the scope of the present invention without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for predicting civil judgment results based on thought chain, characterized in that: The method comprises: Obtaining input information and extracting case elements from the input information; wherein the input information includes: claims, defenses and case facts; Search the knowledge base based on the input information and case elements to obtain historical cases and relevant laws and regulations similar to the input information; Combine the claim and case elements to build the first chain of thought for the purpose of fixing the claim and specifying the claim, call the big model, and obtain the claim; Combining the claims, defenses, case elements, rights requests, historical cases similar to the input information, and relevant laws and regulations, a second thinking chain is constructed to screen the basic norms of rights requests and the basic norms of defense rights, and the large model is called to obtain the basic norms of rights requests, the basic norms of defense rights, and the defense claims and reasons; Combining the basic norms of rights claims and the basic norms of defense rights, constructing a third thinking chain for the purpose of decomposing the constituent elements of the basic norms, calling the big model, and obtaining the constituent elements of the basic norms of rights claims and the constituent elements of the basic norms of defense rights; Combining the claims and reasons of the defense, the basic normative elements of the claims and the basic normative elements of the defense, we build the fourth thinking chain for the purpose of generating the controversial focus of the case, and call the big model to obtain the controversial focus of the case. In combination with the case elements, historical cases similar to the input information, rights claims, claims and reasons of the defense, basic norms for rights claims, basic norms for rights of defense, constituent elements of basic norms for rights claims and constituent elements of basic norms for rights of defense, and the controversial focus of the case, a fifth thinking chain is constructed for the purpose of classifying the elements and generating the judgment results, and the big model is called to obtain the civil judgment prediction results of the input information; wherein, the civil judgment prediction results include: the content of the judgment reasoning and the content of the judgment conclusion.

2. The method according to claim 1, characterized in that The extracting of case elements from the input information includes: Extract litigation requests based on the claims; Extract defense opinions based on the arguments; Extract events based on claims, defenses, and case facts; Extract key information based on claims, defenses and case facts.

3. The method according to claim 2, characterized in that The key information includes: civil acts, right holders, liability holders, types of payments, amounts, quantities, performance holders, performance contents and performance methods.

4. The method according to claim 1, characterized in that: The knowledge base includes: a civil case cognition knowledge base, the civil case cognition knowledge base is composed of a first plain text part and a first vector part, the first plain text part includes a civil historical case and a summary of the civil historical case, case characteristics, and legal content, and the first vector part is generated based on the semantic information of the civil historical case; Based on the input information and case elements, the civil case cognition knowledge base is searched to obtain historical cases similar to the input information, including: Based on the input information, generate a vector representation of the input information, and based on the input information and its vector representation, and in combination with the first plaintext portion and the first vector portion, obtain a first matching result; Based on the case element, a vector representation of the case element is generated, and based on the case element and its vector representation, and in combination with the first plaintext portion and the first vector portion, a second matching result is obtained; Based on the first matching result and the second matching result, historical cases similar to the input information are obtained.

5. The method according to claim 1, characterized in that The knowledge base includes: a civil laws and regulations knowledge base, the civil laws and regulations knowledge base is composed of a second plain text part and a second vector part, the second plain text part includes: the plain text part includes civil law provisions, judicial interpretations and feature tags and themes of the civil law provisions and judicial interpretations, and the second vector part is generated based on the semantic information of the civil law provisions and the judicial interpretations; Search the civil laws and regulations knowledge base based on the input information and case elements to obtain the laws and regulations related to the input information, including: Based on the input information, generate a vector representation of the input information, and based on the input information and its vector representation, and in combination with the second plaintext portion and the second vector portion, obtain a third matching result; Based on the case elements, a vector representation of the case elements is generated, and based on the case elements and the vector representation thereof, and in combination with the second plaintext portion and the second vector portion, a fourth matching result is obtained; Based on the third matching result and the fourth matching result, legal and regulatory provisions related to the input information are obtained.

6. The method according to claim 1, characterized in that The first chain of thinking, the second chain of thinking, the third chain of thinking, the fourth chain of thinking and the fifth chain of thinking also include: detailed reference content that is helpful to complete the corresponding task objectives.

7. The method according to claim 1, characterized in that The training process of the large model includes: Construct training datasets based on data from general and judicial fields; The LoRA fine-tuning method is used on the training data to fine-tune the large model.

8. A civil judgment result prediction system based on thought chain, characterized in that: The system comprises: An information acquisition module is used to acquire input information and extract case elements from the input information; wherein the input information includes: claims, defenses and case facts; The knowledge retrieval module is used to search the knowledge base based on the input information and case elements to obtain historical cases and relevant laws and regulations similar to the input information; The rights request generation module is used to combine the claims and case elements to construct the first thinking chain for the purpose of fixing the claim and specifying the rights request, calling the big model, and obtaining the rights request; The basic norms generation module is used to combine the claims, defenses, case elements, rights requests, historical cases similar to the input information, and relevant laws and regulations to build a second thinking chain for the purpose of screening the basic norms of rights requests and the basic norms of defense rights, and call the big model to obtain the basic norms of rights requests, the basic norms of defense rights, and the defense claims and reasons; The basic normative constituent elements generation module is used to combine the basic norms of the right claim and the basic norms of the right of defense, build a third thinking chain for the purpose of decomposing the basic normative constituent elements, call the big model, and obtain the basic normative constituent elements of the right claim and the basic normative constituent elements of the right of defense; The module for generating the focus of disputes in a case is used to combine the claims, the claims and reasons of the defense, the constituent elements of the basic rules of the claims and the constituent elements of the basic rules of the defense, to construct the fourth thinking chain for the purpose of generating the focus of disputes in the case, and to call the big model to obtain the focus of disputes in the case; The judgment result prediction module is used to combine the case elements, historical cases similar to the input information, rights claims, claims and reasons of the defense, basic norms of rights claims, basic norms of right of defense, constituent elements of basic norms of rights claims and constituent elements of basic norms of right of defense, and the controversial focus of the case, to construct the fifth thinking chain for the purpose of classifying elements and generating judgment results, call the big model, and obtain the civil judgment prediction results of the input information; wherein, the civil judgment prediction results include: the content of the judgment reasoning and the content of the judgment conclusion.

9. An electronic device, characterized in that: The electronic device comprises: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the method for predicting civil judgment results based on thought chain as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement a method for predicting civil judgment results based on a thought chain as described in any one of claims 1 to 7.

Images