Case Evidence Logic Deduction Method Based on Graph Neural Network

Abstract

The present invention discloses a method for logical deduction of case evidence based on graph neural networks. S1. Form a standardized case evidence data set. S2. Form a multi-level case evidence graph. S3. Vectorize the case evidence data nodes and edges of the multi-level case evidence graph, and calculate the weights of the case evidence data nodes. S4. Input the multi-level case evidence graph into a graph neural network to calculate the logical relationship weights between the case evidence data nodes. S5. Identify the case evidence data nodes with low credibility. S6. Based on the logical relationship weights of the case evidence data nodes and the causal reasoning results, use a path search algorithm to calculate the optimal case evidence deduction path, identify the core case fact chain, and generate a case deduction conclusion. The present invention enables different types of case evidence data to be deduced at their respective levels, effectively avoiding information mixing, and improving the accuracy and interpretability of case fact deduction.

Description

Technical Field

[0001] The present invention relates to the technical field of case evidence, and particularly to a method for logical deduction of case evidence based on a graph neural network. Background Art

[0002] With the development of artificial intelligence technology, intelligent analysis tools have gradually been introduced into the judicial field to assist in case trial and evidence reasoning. The logical deduction of case evidence is an important part of judicial reasoning. However, in the traditional process of case analysis and evidence reasoning, the construction and deduction of case facts still mainly rely on the empirical judgments of judges and lawyers, lacking a systematic and intelligent method for constructing an evidence chain. Currently, the evidence analysis of most cases relies on manual retrieval, manual induction, and manual reasoning methods, that is, judges and lawyers need to read case materials one by one, extract key information from a large number of text, image, audio, and time-series evidence, and reason about the case fact chain based on personal experience. Although this method can ensure that legal professionals have an in-depth understanding of the case, due to the diverse types of case evidence and the complex interrelationships, the case analysis process is time-consuming and inefficient.

[0003] In recent years, some legal technology systems have begun to be applied to case analysis and reasoning, mainly relying on legal knowledge graphs and keyword-based case retrieval technologies to assist case analysis. Such technologies have improved the efficiency of case information retrieval to a certain extent, but there are still many problems in case reasoning and fact chain construction: on the one hand, the existing legal information retrieval systems are mainly based on keyword matching, unable to accurately deduce case facts and difficult to identify the logical relationships of key evidence in complex cases; on the other hand, the existing case analysis systems usually lack the ability to model the causal relationships of evidence, making it difficult to reason by integrating different types of evidence, resulting in low accuracy and credibility of case deduction. In addition, existing technologies generally fail to effectively handle the hierarchical relationships between case evidence, unable to distinguish the logical relationships between basic evidence, derived evidence, and legal rules during the deduction process, affecting the scientific nature and interpretability of case reasoning.

[0004] In summary, the existing technologies have obvious deficiencies in the systematization of evidence chain construction, the automation of the reasoning process, the quantitative evaluation of evidence credibility, and the rationality verification of legal rules, and are difficult to meet the intelligent needs of modern judicial case analysis. Summary of the Invention

[0005] An object of the present invention is to propose a method for logical deduction of case evidence based on a graph neural network, which enables different types of case evidence data to be reasoned at their respective levels, effectively avoiding information mixing and improving the accuracy and interpretability of case fact deduction.

[0006] A method for logical deduction of case evidence based on a graph neural network according to an embodiment of the present invention includes the following steps:

[0007] S1. Collect data on case-related information, and preprocess case evidence data of different types to form a standardized case evidence data set;

[0008] S2. Construct a multi-level case evidence graph based on the standardized case evidence data set, set multi-level case evidence nodes, and form a multi-level case evidence graph;

[0009] S3. Vectorize the case evidence data nodes and edges of the multi-level case evidence graph, and calculate the weights of the case evidence data nodes;

[0010] S4. Input the multi-level case evidence graph into a graph neural network to calculate the logical relationship weights between case evidence data nodes;

[0011] S5. Construct a causal reasoning model based on the logical relationship weights of case evidence data nodes, calculate the causal strength between case evidence data nodes, and identify case evidence data nodes with low credibility;

[0012] S6. Based on the logical relationship weights of case evidence data nodes and the causal reasoning results, use a path search algorithm to calculate the optimal case evidence deduction path, identify the core case fact chain, generate a case deduction conclusion, and verify the rationality of the case deduction conclusion in combination with the constraint conditions of legal rule nodes.

[0013] Optionally, S1 includes the following steps:

[0014] S11. Obtain case-related information and construct a case evidence data set. The case evidence data set includes text-based case evidence data, image-based case evidence data, voice-based case evidence data, and time series-based case evidence data. The text-based case evidence data includes witness testimonies, contract agreements, and judgment documents. The voice-based case evidence data includes phone recordings and conversation records. The time series-based case evidence data includes transaction records and location information;

[0015] S12. Perform word segmentation, entity recognition, and semantic parsing on the text-based case evidence data to extract a set of case-related entities. Perform object detection and feature extraction on the image-based case evidence data to obtain a set of key case objects. Perform speech recognition on the voice-based case evidence data and transcribe it into text form to generate a voice-text mapping set. Perform time alignment and trend analysis on the time series-based case evidence data to construct a time series feature matrix;

[0016] S13. Integrate the formatted processing results of case evidence data to form a standardized case evidence data set :

[0017] ;

[0018] Among them, is the set of case text entities, is the set of case image targets, is the set of speech text mappings, is the time series feature matrix.

[0019] Optionally, the S2 includes the following steps:

[0020] S21. For each standardized case evidence data in the standardized case evidence data set , use the multi-modal feature fusion function to generate the evidence node representation :

[0021] ;

[0022] Among them, LayerNorm represents the layer normalization operation, r represents the modality type of the evidence data, taking values from the standardized case evidence data set, is the feature extraction function for modality r, mapping the evidence data d to a preliminary feature vector, is the feature transformation matrix for modality r, is the corresponding bias vector, is the modality weight calculated by the self-attention mechanism;

[0023] Through the mapping function constitute the preliminary set of case evidence graph nodes:

[0024] ;

[0025] and the corresponding preliminary multi-level case evidence graph:

[0026] ;

[0027] S22. Divide the preliminary node set into different levels according to its feature distribution:

[0028] ;

[0029] Among them, is any evidence node, l represents the level type of the case evidence graph, taking values from the basic case evidence layer base, the derived case evidence layer infer, and the legal rule layer law, is the predefined level prototype vector, reflecting the feature center that each layer should have;

[0030] Then the node sets at each level are defined as follows:

[0031] ;

[0032] S23. Based on the association relationships between the case evidence data nodes, establish edges between the basic case evidence layer and the derived case evidence layer . For any basic node and the derived node , set its edge weight as:

[0033] ;

[0034] where represents the cosine similarity between two nodes, is the Gaussian kernel scale parameter used to control the similarity attenuation, is the preset similarity threshold, is the indicator function that takes the value 1 when the condition is satisfied and 0 otherwise;

[0035] Thus, an edge set is formed:

[0036] ;

[0037] S24. Based on the relevance between the case evidence data nodes and the legal rules, establish edges between the nodes in the basic and derived layers and the nodes in the legal rule layer. For any evidence node and the legal rule node , set its edge weight as:

[0038] ;

[0039] where represents the Sigmoid function used to map the similarity to [0, 1], represents the inner product of two vectors;

[0040] Thus, an edge set is formed:

[0041] ;

[0042] where is the threshold for legal rule association;

[0043] S25. Merge the edge set between the basic and derived layers constructed in step S23 and the edge set between the case evidence and the legal rules constructed in step S24 to form the complete edge set of the multi-level case evidence graph:

[0044] ;

[0045] The fully constructed multi-level case evidence graph is as follows:

[0046] 。

[0047] Optionally, S3 includes the following steps:

[0048] S31. Perform feature embedding on each case evidence data node based on the multi-level case evidence graph to generate an initial feature vector :

[0049] ;

[0050] Among them, represents the type embedding of the case evidence data node, corresponding to the basic case evidence, derived case evidence, or legal rule, represents the content embedding of the case evidence data node, based on the feature extraction results of case evidence data text, image, voice, or time series, represents the relevance embedding of the case evidence data node, calculated based on the relevance between the case evidence data and other case evidence data;

[0051] S32. Perform vector representation on each edge based on the edge set and calculate the edge feature representation :

[0052] ;

[0053] Among them, are the initial feature vectors of case evidence data nodes u and v respectively, is the edge feature calculation function, using weighted concatenation or non-linear transformation;

[0054] S33. Based on the embedding vectors of the case evidence data nodes and the vector representation of the edges, construct a structured case evidence representation matrix:

[0055] ;

[0056] Among them, is the feature matrix of the case evidence data node, and each row corresponds to the initial feature vector of a case evidence data node. N is the total number of case evidence data nodes, and d is the dimension of the feature vector;

[0057] Construct an adjacency matrix based on the weight information of the edges:

[0058] ;

[0059] Among them, is the adjacency matrix of the multi-level case evidence graph, representing the association strength between case evidence data nodes;

[0060] S34. Feature matrix based on case evidence data nodes and adjacency matrix , calculate the global importance score of case evidence data nodes:

[0061] ;

[0062] wherein, is the set of neighbor nodes of node v, is the attention weight, indicating the influence of node u on node v:

[0063] ;

[0064] wherein, a is a vector of trainable parameters, is the activation function;

[0065] Calculate the normalized weight of case evidence data nodes:

[0066] .

[0067] Optionally, the S4 includes the following steps:

[0068] S41. Based on the adjacency matrix, calculate the information aggregation between case evidence data nodes through the graph convolution mechanism:

[0069] ;

[0070] wherein, is the updated feature representation of case evidence data node v in the (l + 1)-th layer of the graph neural network, is the edge weight between case evidence data nodes u and v, is the normalization factor of case evidence data node v, is the trainable parameter matrix of the l-th layer, is the bias vector;

[0071] S42. Calculate the attention weights between case evidence data nodes based on the attention mechanism ;

[0072] S43. Define the hierarchical structure relationship of case evidence data nodes according to the hierarchical division, and perform feature propagation to calculate and deduce the updated features of case evidence data nodes:

[0073] ;

[0074] wherein, is the deduced updated feature vector of case evidence data node in the (l + 1)-th layer, and are the feature vectors of the basic case evidence data node and the legal rule node at the l-th layer, respectively, and are the attention coefficients between the basic case evidence data node and the derived case evidence data node, and between the legal rule node and the derived case evidence data node, respectively;

[0075] S44. Calculate the logical relationship weight between case evidence data nodes based on the final feature representation of the case evidence data nodes:

[0076] ;

[0077] where, is the logical relationship weight between case evidence data nodes u and v, reflecting the association strength between the two nodes, and are the feature representations of case evidence data nodes u and v at the final layer L, is the logical relationship mapping matrix, is the bias term.

[0078] Optionally, the S5 includes the following steps:

[0079] S51. Define the structural equation of the causal inference model based on the logical relationship weight matrix between case evidence data nodes:

[0080] ;

[0081] where, is the causal effect representation of case evidence data node v, is the causal coefficient in the causal inference model, is the feature representation of case evidence data node u at the final layer L, is the noise term, representing the random influence of the existence of case evidence data node v;

[0082] S52. Calculate the causal strength of case evidence data node u on node v based on the causal inference model:

[0083] ;

[0084] where, is the causal strength matrix of case evidence data node u on node v, reflecting the causal relationship strength between the two, is the adjustment coefficient of the causal strength, used to control the non-linear mapping of the causal strength;

[0085] S53. Calculate the credibility score of the case evidence data node based on the causal strength ;

[0086] S54. Update the weight of the case evidence data node based on the credibility score:

[0087] ;

[0088] Among them, is the updated weight of the case evidence data node v, is the adjustment coefficient, which controls the influence degree of the credibility score on the weight;

[0089] S55. Set the credibility threshold , and identify the set of case evidence data nodes with low credibility:

[0090] ;

[0091] Among them, is the set of case evidence data nodes with low credibility, indicating the case evidence data nodes with incomplete or incorrect information.

[0092] Optionally, the S6 includes the following steps:

[0093] S61. Define the path search weight matrix between case evidence data nodes based on the logical relationship weight matrix and the causal strength matrix:

[0094] ;

[0095] Among them, represents the path weight between the case evidence data node u and v, is the logical relationship weight between the case evidence data node u and v, is the adjustment factor, which controls the balance between the logical relationship weight and the causal strength;

[0096] S62. Based on the path weight matrix P between case evidence data nodes, set the initial case evidence data node s and the case conclusion node t, and solve the optimal case evidence deduction path through the shortest path search algorithm:

[0097] ;

[0098] Among them, is the optimal path of the case evidence deduction, is the set of all possible case evidence deduction paths, represents the path the sum of the path weights of all edges on;

[0099] S63. Calculate the key nodes on the path based on the optimal case evidence deduction path, and define the case core fact chain:

[0100] ;

[0101] Among them, is the set of key case evidence data nodes on the core case fact chain, is the threshold of causal strength, used to screen case evidence data nodes with high causal influence;

[0102] S64. Define the case deduction conclusion based on the core case fact chain ;

[0103] S65. Calculate the legal rule consistency score based on the legal rule node and the case deduction conclusion:

[0104] ;

[0105] Among them, is the consistency score between the case deduction conclusion and the legal rule, is the logical relationship weight between the case evidence data node v on the core case fact chain and the legal rule node k;

[0106] Based on the legal rule consistency score set the rationality judgment standard , if , then judge that the case deduction conclusion is reasonable, otherwise readjust the case evidence deduction path for iterative optimization.

[0107] The beneficial effects of the present invention are:

[0108] (1) By constructing a multi-level case evidence graph, the present invention divides case evidence data into a basic case evidence layer, a derived case evidence layer, and a legal rule layer, and adopts an adaptive feature embedding method to generate a structured case evidence representation matrix based on the type, content, and relevance of evidence data, enabling case evidence data to be stored and modeled in a unified graph structure. Using the graph structure to explicitly represent the causal relationship between case evidence data makes the reasoning path clearer, can effectively solve the problems of incomplete evidence chains and unclear logical relationships in traditional case analysis, and at the same time enables different types of case evidence data to be reasoned in their respective levels, effectively avoiding information mixing and improving the accuracy and interpretability of case fact deduction.

[0109] (2) The present invention uses a graph neural network for logical reasoning of case evidence, and combines an attention mechanism to optimize the influence of the core evidence nodes of the case, enabling the reasoning model to adaptively learn the high-order relationships between case evidence data. In the information propagation process, a feature propagation method based on hierarchical attention is adopted, enabling the derived case evidence data nodes to obtain key information from the basic case evidence data nodes and legal rule nodes, thereby enhancing the rationality and reliability of the derived evidence in the case reasoning process. The graph neural network reasoning method can automatically learn the internal correlation of case evidence data, improve the automation degree of evidence logical analysis, and reduce the influence of human subjective factors on the case reasoning result.

[0110] (3) The present invention constructs a causal reasoning model based on the logical relationship weight and causal strength of case evidence data, calculates the credibility of case evidence data nodes, and uses a dynamic weight adjustment method to optimize the case deduction path. For the rationality verification of case deduction, a legal rule consistency scoring mechanism is introduced to ensure that the case reasoning result conforms to relevant legal norms. Compared with the existing case reasoning methods based on statistical analysis, the causal reasoning method of the present invention can effectively identify the causal chain of case evidence data and find the optimal deduction path through a path search algorithm, thereby improving the reliability of the core fact chain of the case. Brief Description of the Drawings

[0111] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0112] Figure 1 is a flowchart of a method for logical deduction of case evidence based on a graph neural network proposed by the present invention. Detailed Embodiments

[0113] Now, the present invention will be further described in detail with reference to the drawings. These drawings are all simplified schematic diagrams, only showing the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0114] Refer to Figure 1 , a method for logical deduction of case evidence based on a graph neural network, includes the following steps:

[0115] S1. Collect data on case-related information and preprocess different types of case evidence data to form a standardized case evidence data set;

[0116] S2. Construct a multi-level case evidence graph based on the standardized case evidence data set, set multi-level case evidence nodes, and form a multi-level case evidence graph;

[0117] S3. Vectorize the case evidence data nodes and edges in the multi-level case evidence graph, and calculate the weights of the case evidence data nodes;

[0118] S4. Input the multi-level case evidence graph into a graph neural network to calculate the logical relationship weights between the case evidence data nodes;

[0119] S5. Construct a causal inference model based on the logical relationship weights of the case evidence data nodes, calculate the causal strength between the case evidence data nodes, and identify the case evidence data nodes with low credibility;

[0120] S6. Based on the logical relationship weights of the case evidence data nodes and the causal inference results, use a path search algorithm to calculate the optimal case evidence deduction path, identify the core case fact chain, generate a case deduction conclusion, and verify the rationality of the case deduction conclusion in combination with the constraint conditions of the legal rule nodes.

[0121] In this embodiment, S1 includes the following steps:

[0122] S11. Obtain case-related information and construct a case evidence dataset, which includes text-based case evidence data, image-based case evidence data, voice-based case evidence data, and time-series-based case evidence data. The text-based case evidence data includes witness testimonies, contract agreements, and judgment documents. The voice-based case evidence data includes phone recordings and conversation records. The time-series-based case evidence data includes transaction records and location information;

[0123] S12. Perform word segmentation, entity recognition, and semantic parsing on the text-based case evidence data to extract the entity set related to the case. Perform object detection and feature extraction on the image-based case evidence data to obtain the key object set of the case. Perform speech recognition on the voice-based case evidence data and transcribe it into text form to generate a voice-text mapping set. Perform time alignment and trend analysis on the time-series-based case evidence data to construct a time-series feature matrix;

[0124] S13. Integrate the formatted processing results of the case evidence data to form a standardized case evidence dataset :

[0125] ;

[0126] Among them, is the case text entity set, is the case image object set, is the voice-text mapping set, is the time-series feature matrix.

[0127] In this embodiment, S2 includes the following steps:

[0128] S21. For each piece of standardized case evidence data in the standardized case evidence data set, , use the multi-modal feature fusion function to generate the evidence node representation :

[0129] ;

[0130] where LayerNorm represents the layer normalization operation, r represents the modality type of the evidence data, taking values from the standardized case evidence data set, is the feature extraction function for modality r, mapping the evidence data d to a preliminary feature vector, is the feature transformation matrix for modality r, is the corresponding bias vector, is the modality weight calculated by the self-attention mechanism;

[0131] Through the mapping function constitute a preliminary set of nodes for the case evidence graph:

[0132] ;

[0133] And the corresponding preliminary multi-level case evidence graph:

[0134] ;

[0135] S22. Divide the preliminary node set into different levels according to its feature distribution:

[0136] ;

[0137] where, is any evidence node, l represents the level type of the case evidence graph, taking values from the basic case evidence layer base, the derived case evidence layer infer, and the legal rule layer law, is the predefined level prototype vector, reflecting the feature center that each layer should have;

[0138] Then the node sets of each level are respectively defined as:

[0139] ;

[0140] S23. Based on the association relationship between the case evidence data nodes, establish an edge between the basic case evidence layer and the derived case evidence layer . For any basic node and the derived node , set its edge weight as:

[0141] ;

[0142] Among them, represents the cosine similarity between two nodes, is the Gaussian kernel scale parameter used to control the similarity attenuation, is a preset similarity threshold, is an indicator function that takes the value 1 when the condition is satisfied and 0 otherwise;

[0143] Thus, an edge set is formed:

[0144] ;

[0145] S24. Establish an edge between the basic and derived layer nodes and the legal rule layer nodes based on the relevance between the case evidence data nodes and the legal rules. For any evidence node and the legal rule node , set its edge weight as:

[0146] ;

[0147] Among them, represents the Sigmoid function used to map the similarity to [0, 1], represents the inner product of two vectors;

[0148] Thus, an edge set is formed:

[0149] ;

[0150] Among them, is the threshold for legal rule association;

[0151] S25. Merge the edge set between the basic and derived layers constructed in step S23 and the edge set between the case evidence and the legal rules constructed in step S24 to form the complete edge set of the multi-level case evidence graph:

[0152] ;

[0153] The completely constructed multi-level case evidence graph is:

[0154] .

[0155] In this embodiment, S3 includes the following steps:

[0156] S31. Perform feature embedding on each case evidence data node based on the multi-level case evidence graph to generate an initial feature vector :

[0157] ;

[0158] Among them, represents the type embedding of the case evidence data node, corresponding to the basic case evidence, derived case evidence, or legal rule, represents the content embedding of the case evidence data node, which is based on the feature extraction results of case evidence data text, image, voice, or time series, represents the relevance embedding of the case evidence data node, which is calculated based on the relevance between the case evidence data and other case evidence data;

[0159] S32. Vectorize each edge based on the edge set and calculate the edge feature representation :

[0160] ;

[0161] Among them, are the initial feature vectors of the case evidence data nodes u and v respectively, is the edge feature calculation function, which uses weighted splicing or non-linear transformation;

[0162] S33. Based on the embedding vectors of the case evidence data nodes and the vectorized representation of the edges, construct a structured case evidence representation matrix:

[0163] ;

[0164] Among them, is the feature matrix of the case evidence data node, where each row corresponds to the initial feature vector of a case evidence data node, N is the total number of case evidence data nodes, and d is the dimension of the feature vector;

[0165] Construct an adjacency matrix based on the weight information of the edges:

[0166] ;

[0167] Among them, is the adjacency matrix of the multi-level case evidence graph, representing the association strength between case evidence data nodes;

[0168] S34. Based on the feature matrix of the case evidence data node and the adjacency matrix , calculate the global importance score of the case evidence data node:

[0169] ;

[0170] Among them, is the set of neighbor nodes of node v, is the attention weight, representing the influence of node u on node v:

[0171] ;

[0172] Among them, a is a vector of trainable parameters, is an activation function;

[0173] Calculate the normalized weights of the case evidence data nodes:

[0174] .

[0175] In this embodiment, S4 includes the following steps:

[0176] S41. Based on the adjacency matrix, calculate the information aggregation between the case evidence data nodes through the graph convolution mechanism:

[0177] ;

[0178] Among them, is the updated feature representation of the case evidence data node v in the (l + 1)-th layer of the graph neural network, is the edge weight between the case evidence data nodes u and v, is the normalization factor of the case evidence data node v, is the trainable parameter matrix of the l-th layer, is the bias vector;

[0179] S42. Calculate the attention weights between the case evidence data nodes based on the attention mechanism ;

[0180] S43. Define the hierarchical structure relationship of the case evidence data nodes according to the hierarchical division, and perform feature propagation to calculate and deduce the updated features of the case evidence data nodes:

[0181] ;

[0182] Among them, is the updated feature vector of the deduced case evidence data node in the (l + 1)-th layer, and are the feature vectors of the basic case evidence data node and the legal rule node in the l-th layer respectively, and are the attention coefficients between the basic case evidence data node and the deduced case evidence data node, and between the legal rule node and the deduced case evidence data node respectively;

[0183] S44. Calculate the logical relationship weights between the case evidence data nodes based on the final feature representation of the case evidence data nodes:

[0184] ;

[0185] Among them, is the logical relationship weight between case evidence data nodes u and v, reflecting the association strength between the two nodes, and are the feature representations of case evidence data nodes u and v at the final layer L, is the logical relationship mapping matrix, is the bias term.

[0186] In this embodiment, S5 includes the following steps:

[0187] S51. Define the structural equation of the causal inference model based on the logical relationship weight matrix between case evidence data nodes:

[0188] ;

[0189] Among them, is the causal effect representation of case evidence data node v, is the causal coefficient in the causal inference model, is the feature representation of case evidence data node u at the final layer L, is the noise term, indicating the random influence of the existence of case evidence data node v;

[0190] S52. Calculate the causal strength of case evidence data node u on node v based on the causal inference model:

[0191] ;

[0192] Among them, is the causal strength matrix of case evidence data node u on node v, reflecting the causal relationship strength between the two, is the adjustment coefficient of the causal strength, used to control the non-linear mapping of the causal strength;

[0193] S53. Calculate the credibility score of case evidence data nodes based on the causal strength ;

[0194] S54. Update the weight of case evidence data nodes based on the credibility score

[0195] ;

[0196] Among them, is the updated weight of case evidence data node v, is the adjustment coefficient, controlling the influence degree of the credibility score on the weight;

[0197] S55. Set the credibility threshold , identify the set of case evidence data nodes with low credibility:

[0198] ;

[0199] Among them, is the set of case evidence data nodes with low credibility, indicating case evidence data nodes with incomplete or incorrect information.

[0200] In this embodiment, S6 includes the following steps:

[0201] S61. Define the path search weight matrix between case evidence data nodes based on the logical relationship weight matrix and the causal strength matrix:

[0202] ;

[0203] Among them, represents the path weight between case evidence data nodes u and v, is the logical relationship weight between case evidence data nodes u and v, is the adjustment factor, controlling the balance between the logical relationship weight and the causal strength;

[0204] S62. Based on the path weight matrix P between case evidence data nodes, set the initial case evidence data node s and the case conclusion node t, and solve the optimal case evidence deduction path through the shortest path search algorithm:

[0205] ;

[0206] Among them, is the optimal path of case evidence deduction, is the set of all possible case evidence deduction paths, represents the path the sum of the path weights of all edges on;

[0207] S63. Calculate the key nodes on the path based on the optimal case evidence deduction path, and define the case core fact chain:

[0208] ;

[0209] Among them, is the set of key case evidence data nodes on the case core fact chain, is the threshold of causal strength, used to screen case evidence data nodes with high causal influence;

[0210] S64. Define the case deduction conclusion based on the case core fact chain ;

[0211] S65. Calculate the legal rule consistency score based on legal rule nodes and case deduction conclusions:

[0212] ;

[0213] Among them, is the consistency score between the case deduction conclusion and the legal rule, is the logical relationship weight between the case evidence data node v on the core case fact chain and the legal rule node k;

[0214] Based on the legal rule consistency score Set the rationality judgment standard , if , then determine that the case deduction conclusion is reasonable, otherwise readjust the case evidence deduction path for iterative optimization.

[0215] The present invention divides case evidence data into a basic case evidence layer, a deduced case evidence layer, and a legal rule layer by constructing a multi-level case evidence graph, and adopts an adaptive feature embedding method to generate a structured case evidence representation matrix based on the type, content, and relevance of evidence data, enabling case evidence data to be stored and modeled in a unified graph structure. Using the graph structure to explicitly represent the causal relationship between case evidence data makes the reasoning path clearer, can effectively solve the problems of incomplete evidence chains and unclear logical relationships in traditional case analysis, and at the same time enables different types of case evidence data to be reasoned in their respective levels, effectively avoiding information mixing and improving the accuracy and interpretability of case fact deduction.

[0216] The present invention uses a graph neural network for case evidence logical reasoning and combines an attention mechanism to optimize the influence of core case evidence nodes, enabling the reasoning model to adaptively learn the high-order relationships between case evidence data. In the information propagation process, a feature propagation method based on hierarchical attention is adopted so that the deduced case evidence data nodes can obtain key information from the basic case evidence data nodes and legal rule nodes, thereby enhancing the rationality and reliability of the deduced evidence in the case reasoning process. The graph neural network reasoning method can automatically learn the internal relevance of case evidence data, improve the automation degree of evidence logical analysis, and reduce the influence of human subjective factors on the case reasoning result.

[0217] The present invention constructs a causal reasoning model based on the logical relationship weights and causal strengths of case evidence data, calculates the credibility of case evidence data nodes, and uses a dynamic weight adjustment method to optimize the case deduction path. For the rationality verification of case deduction, a legal rule consistency scoring mechanism is introduced to ensure that the case reasoning results comply with relevant legal norms. Compared with the existing case reasoning methods based on statistical analysis, the causal reasoning method of the present invention can effectively identify the causal chain of case evidence data and find the optimal deduction path through a path search algorithm, thereby improving the reliability of the core fact chain of the case.

[0218] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and all should be covered by the protection scope of the present invention.

Claims

1. A case evidence logic deduction method based on graph neural network, characterized in that: The steps include: S1. Collect case-related information and pre-process different types of case evidence data to form a standardized case evidence data set; S2. Based on the standardized case evidence data set, multi-level case evidence nodes are set to form a multi-level case evidence graph; S3, vectorizing the case evidence data nodes and edges of the multi-level case evidence graph, and calculating the weights of the case evidence data nodes; S4, inputting the multi-level case evidence graph into the graph neural network to calculate the logical relationship weights between the case evidence data nodes; S5. Construct a causal reasoning model based on the logical relationship weights of case evidence data nodes, calculate the causal strength between case evidence data nodes, and identify case evidence data nodes with low credibility; S6. Based on the logical relationship weights and causal reasoning results of the case evidence data nodes, a path search algorithm is used to calculate the optimal case evidence deduction path, identify the core fact chain of the case, generate the case deduction conclusion, and verify the rationality of the case deduction conclusion in combination with the constraints of the legal rule nodes; The S2 comprises the following steps: S21. For each standardized case evidence data in the standardized case evidence data set , using multimodal feature fusion function to generate evidence node representation : ； Among them, LayerNorm represents the layer normalization operation, r represents the modal type of evidence data, which is taken from the standardized case evidence dataset. is the feature extraction function for modality r, mapping the evidence data d into a preliminary feature vector, is the characteristic transformation matrix of mode r, is the corresponding bias vector, The modal weight calculated by the self-attention mechanism; Through the mapping function The node set that constitutes the preliminary case evidence graph is: ； And the corresponding preliminary multi-level case evidence map: ； S22, the initial node set According to their characteristic distribution, they are divided into different levels: ； in, is any evidence node, l represents the hierarchical type of the case evidence graph, which is taken from the base case evidence layer, the inferred case evidence layer, and the legal rule layer, It is a predefined hierarchical prototype vector, reflecting the feature center that each layer should have; Then the node sets of each level are defined as: ； S23. Based on the relationship between case evidence data nodes, and deriving the case evidence layer Establish an edge between any basic node With the inference node , set its edge weight to: ； in, represents the cosine similarity between two nodes, is the Gaussian kernel scale parameter, which is used to control the similarity attenuation. is the preset similarity threshold, is an indicator function, which takes the value 1 when the condition is met, otherwise it takes the value 0; This forms the edge set: ； S24. Based on the correlation between the case evidence data nodes and the legal rules, establish edges between the basic and derivation layer nodes and the legal rule layer nodes. Legal rule nodes , set its edge weight to: ； in, represents the Sigmoid function, which is used to map the similarity to [0,1], represents the inner product of two vectors; This forms the edge set: ； in, Thresholds associated with legal rules; S25, merging the edge set between the base and the derivation layer constructed in step S23 with the edge set between the case evidence and the legal rules constructed in step S24 to form a complete edge set of the multi-level case evidence graph: ； The fully constructed multi-layered case evidence diagram is as follows: 。 2. According to claim 1, a case evidence logic deduction method based on graph neural network is characterized in that: The S1 comprises the following steps: S11. Obtain case-related information and construct a case evidence data set, which includes text case evidence data, image case evidence data, voice case evidence data, and time series case evidence data. The text case evidence data includes witness testimony, contract agreements, and judgment documents, the voice case evidence data includes telephone recordings and conversation records, and the time series case evidence data includes transaction records and location information. S12. Perform word segmentation, entity recognition and semantic analysis on text case evidence data, extract case-related entity sets, perform target detection and feature extraction on image case evidence data, obtain case key target sets, perform speech recognition on voice case evidence data, transcribe it into text form, generate speech-text mapping sets, perform time alignment and trend analysis on time series case evidence data, and construct a time series feature matrix; S13. Integrate the formatting results of case evidence data to form a standardized case evidence data set : ； in, is the case text entity set, is the case image target set, is a speech-to-text mapping collection, is the time series feature matrix.

3. According to claim 1, a case evidence logic deduction method based on graph neural network is characterized in that: The S3 comprises the following steps: S31. Based on the multi-level case evidence graph, feature embedding is performed on each case evidence data node to generate an initial feature vector : ； in, Indicates the type embedding of case evidence data nodes, corresponding to basic case evidence, derived case evidence or legal rules, Represents the content embedding of case evidence data nodes, based on the feature extraction results of case evidence data text, image, speech or time series, Relationship embedding representing case evidence data nodes, which is calculated based on the correlation between case evidence data and other case evidence data; S32. Vectorize each edge based on the edge set and calculate the edge feature representation : ； in, are the initial feature vectors of case evidence data nodes u and v respectively, Calculate the function for edge features, using weighted concatenation or nonlinear transformation; S33. Based on the embedding vectors of case evidence data nodes and the vectorized representation of edges, a structured case evidence representation matrix is ​​constructed: ； in, is the feature matrix of the case evidence data node, each row corresponds to the initial feature vector of a case evidence data node, N is the total number of case evidence data nodes, and d is the dimension of the feature vector; Construct an adjacency matrix based on edge weight information: ； in, is the adjacency matrix of the multi-level case evidence graph, which represents the association strength between the case evidence data nodes; S34. Feature matrix based on case evidence data nodes and the adjacency matrix , calculate the global importance score of the case evidence data node: ； in, is the set of neighbor nodes of node v, is the attention weight, which indicates the influence of node u on node v: ； Among them, a is the trainable parameter vector, is the activation function; Calculate the normalized weight of the case evidence data node: 。 4. According to claim 1, a case evidence logic deduction method based on graph neural network is characterized in that: The S4 comprises the following steps: S41. Based on the adjacency matrix, the information aggregation between case evidence data nodes is calculated through the graph convolution mechanism: ； in, is the updated feature representation of the case evidence data node v in the l+1th layer of the graph neural network, is the edge weight between case evidence data nodes u and v, is the normalization factor of the case evidence data node v, is the trainable parameter matrix of layer l, is the bias vector; S42. Calculate the attention weights between case evidence data nodes based on the attention mechanism ; S43. Define the hierarchical relationship of the case evidence data nodes according to the hierarchical division, perform feature propagation, and calculate and derive the update features of the case evidence data nodes: ； in, To derive case evidence data nodes The updated feature vector at layer l+1 is, and are the feature vectors of the basic case evidence data node and the legal rule node at the lth layer, and They are the attention coefficients between the basic case evidence data node and the derived case evidence data node, and between the legal rule node and the derived case evidence data node; S44. Calculate the logical relationship weights between the case evidence data nodes based on the final feature representation of the case evidence data nodes: ； in, is the logical relationship weight between case evidence data nodes u and v, reflecting the strength of association between the two nodes. and is the feature representation of case evidence data nodes u and v at the final layer L, is the logical relationship mapping matrix, is the bias term.

5. According to claim 1, a case evidence logic deduction method based on graph neural network is characterized in that: The S5 comprises the following steps: S51. Define the structural equation of the causal reasoning model based on the logical relationship weight matrix between case evidence data nodes: ； in, is the causal effect representation of the case evidence data node v, is the causal coefficient in the causal inference model, is the feature representation of the case evidence data node u at the final layer L, is the noise term, which indicates the random influence of the case evidence data node v; S52. Calculate the causal strength of case evidence data node u to node v based on the causal reasoning model: ； in, is the causal strength matrix of case evidence data node u to node v, reflecting the strength of the causal relationship between the two. is the adjustment coefficient of causal strength, which is used to control the nonlinear mapping of causal strength; S53. Calculate the credibility score of case evidence data nodes based on causal strength ; S54. Update the weight of the case evidence data node based on the credibility score: ； in, is the updated weight of the case evidence data node v, To adjust the coefficient, control the influence of the credibility score on the weight; S55. Setting credibility threshold , identify the set of case evidence data nodes with low credibility: ； in, It is a set of case evidence data nodes with low credibility, indicating that there are case evidence data nodes with incomplete or erroneous information.

6. The case evidence logic deduction method based on graph neural network according to claim 1 is characterized in that: The S6 comprises the following steps: S61. Define the path search weight matrix between case evidence data nodes based on the logical relationship weight matrix and the causal strength matrix: ； in, represents the path weight between case evidence data nodes u and v, is the logical relationship weight between case evidence data nodes u and v, As a regulatory factor, it controls the balance between the weight of the logical relationship and the causal strength; S62. Based on the path weight matrix P between the case evidence data nodes, the case initial evidence data node s and the case conclusion node t are set, and the optimal case evidence deduction path is solved by the shortest path search algorithm: ； in, The optimal path for case evidence deduction, A set of deduction paths for all possible case evidence, Indicates the path The sum of the path weights of all edges above; S63. Based on the optimal case evidence deduction path, calculate the key nodes on the path and define the core fact chain of the case: ； in, It is a collection of key case evidence data nodes on the core fact chain of the case. is the threshold of causal strength, which is used to screen case evidence data nodes with high causal influence; S64, based on the core facts of the case Defining Case Conclusions ; S65. Calculate the legal rule consistency score based on the legal rule nodes and case deduction conclusions: ； in, Score the consistency between the case deduction conclusion and the legal rules, is the weight of the logical relationship between the case evidence data node v and the legal rule node k on the case core fact chain; Based on the legal rule consistency score Setting reasonableness criteria ,like , then the case deduction conclusion is judged to be reasonable, otherwise the case evidence deduction path will be readjusted and iterative optimization will be performed.

Images