A method for associating multiple evidence in case records based on knowledge graph

Abstract

The present invention discloses a method for associating multiple evidences in case records based on knowledge graph, S1, generating a basic data set for semantic analysis; S2, performing named entity recognition on the basic data set to extract key entities in the case records; S3, performing relationship extraction on the extracted structured data to identify the temporal relationship, spatial relationship, causal relationship between different entities and the logical association between evidences; S4, constructing a weighted hypergraph structure, using the weighted hypergraph to model and represent the relationship between multiple entities; S5, mapping the nodes and hyperedge information in the weighted hypergraph to the case record knowledge graph; S6, using graph query and hypergraph analysis technology, jointly analyzing the constructed case record knowledge graph and the weighted hypergraph to generate evidence association analysis results; S7, generating a case analysis report based on the evidence association analysis results. The present invention significantly improves the automation level and processing efficiency of case analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of case records, and in particular to a method for associating multiple pieces of evidence in case records based on a knowledge graph. Background Art

[0002] With the development of information technology, a large amount of judicial case data is stored in digital form, especially case records, which usually exist in unstructured text form. Case records generally contain the core information of the case, including people, places, time, events and evidence. However, due to its unstructured characteristics, the process of processing and analyzing case records is often very complicated and time-consuming. When conducting case analysis, judicial personnel need to extract relevant information from massive case records and establish connections between various types of evidence. This not only consumes a lot of manpower, but is also prone to errors or omissions due to human factors, affecting the efficiency of case trial and decision-making.

[0003] In the existing technology, the judicial system has tried to automate the processing of case records through natural language processing technology, including using named entity recognition technology to extract key information from the text, or using rule engines and machine learning models to identify simple associations between evidence. However, these technical solutions have obvious limitations. First, the accuracy of existing natural language processing technology is often not ideal when dealing with complex legal terms and contexts, especially when dealing with complex relationships across texts, which is prone to misidentification or information omissions. Second, the existing technology focuses more on the extraction and analysis of a single piece of evidence, and lacks the ability to comprehensively analyze the complex associations between multiple pieces of evidence. This makes it difficult to construct a complete case progress path when dealing with cases involving multiple witnesses, multiple physical evidence or complex timelines, resulting in a lack of comprehensive basis for judicial personnel to make decisions. In addition, existing systems usually use simple graph structures to represent the relationship between evidence, which is difficult to deal with complex scenarios with multiple entities and multiple relationships. Especially in large-scale case data analysis, the efficiency and effectiveness of this method are not ideal.

[0004] In response to the above-mentioned deficiencies in the prior art, this application proposes a method for associating multiple evidence in case records based on knowledge graphs, aiming to solve the problems of low information processing efficiency and insufficient analysis of multiple evidence associations in the prior art, and to provide judicial personnel with more comprehensive and reliable decision-making support. Summary of the invention

[0005] One purpose of the present invention is to propose a method for associating multiple pieces of evidence in case records based on a knowledge graph. The present invention can efficiently and accurately identify key evidence and the relationships between them in large-scale case data, significantly improving the degree of automation and processing efficiency of case analysis.

[0006] A method for associating multiple pieces of evidence in a case record based on a knowledge graph according to an embodiment of the present invention includes the following steps:

[0007] S1. Use natural language processing technology to pre-process unstructured case records, including word segmentation, part-of-speech tagging, and syntactic analysis of the text to generate a basic data set for semantic analysis;

[0008] S2. Perform named entity recognition on the basic data set to extract key entities in the case transcripts, including people, places, time, events, and evidence information, and mark the key entities as structured data;

[0009] S3, extract relationships from the extracted structured data to identify the temporal, spatial, causal relationships between different entities and the logical associations between evidence;

[0010] S4. construct a weighted hypergraph structure, and use the weighted hypergraph to model and represent the relationship between multiple entities, wherein the weighted hypergraph nodes represent key entities in the case record, the hyperedges represent the association relationship between different entities, and the weights of the hyperedges are used to describe the strength or importance of the association relationship;

[0011] S5, mapping the node and hyperedge information in the weighted hypergraph to the case record knowledge graph;

[0012] S6. Use graph query and hypergraph analysis technology to jointly analyze the constructed case record knowledge graph and weighted hypergraph, automatically identify and deduce the association relationship between multiple evidences in the case, and generate evidence association analysis results;

[0013] S7. Generate a case analysis report based on the results of the evidence correlation analysis.

[0014] Optionally, the S1 includes the following steps:

[0015] S11, performing word segmentation processing on the text data of the unstructured case record, splitting the continuous original text data T of the case record into word sequences W;

[0016] S12, perform part-of-speech tagging on the word sequence W to generate a part-of-speech tagging sequence P;

[0017] S13, perform syntactic analysis on the part-of-speech tag sequence P to generate a syntactic dependency tree D:

[0018] D = (N, E);

[0019] Where N represents the node set in syntactic analysis, each node corresponds to a word and its part-of-speech tag, E represents the edge set of syntactic dependency, and each edge represents the grammatical dependency between words;

[0020] S14. Extract the subject-verb-object structure and modification relationship in the sentence based on the syntactic dependency tree D to form a structured semantic fragment ;

[0021] S15. Semantic fragments Constructed as a basic dataset for semantic analysis:

[0022] ;

[0023] Among them, the basic data set Including semantic fragments , part-of-speech tag sequence P and word sequence W.

[0024] Optionally, S2 includes the following steps:

[0025] S21. Basic data sets from semantic analysis Each semantic fragment Perform named entity recognition to identify key entities E in the case transcript, including people, places, time, events, and evidence information, and generate an entity set:

[0026] ;

[0027] in, represents a single key entity in the named entity recognition process, and k is the number of key entities identified;

[0028] S22. For each key entity Mark the attribute as structured data ;

[0029] S23, the marked structured data With the original basic data set Perform association to generate an extended structured dataset :

[0030] .

[0031] Optionally, S3 includes the following steps:

[0032] S31. From the extended structured dataset Extract each key entity and associated attribute information ;

[0033] S32. Extracting key entities Analyze the associations with other entities to identify the temporal relationships between entities , time relationship Defined as the order of entities on the timeline;

[0034] S33. Identify spatial relationships between entities , where the spatial relationship Defined as the relative relationship of entities in space;

[0035] S34. Identify causal relationships between entities ,causation Defined as a key entity The occurrence of Occurrence of;

[0036] S35. Identify logical associations between entities , logical association Defined as the existence of interdependent or contradictory relationships between pieces of evidence;

[0037] S36: Combine the time relationship, spatial relationship, causal relationship and logical relationship identified in steps S32 to S35 to conduct a comprehensive correlation analysis of multiple pieces of evidence, identify the relationship between multiple pieces of evidence in the case record, and integrate them into the extended structured data set. Generate a dataset containing entities and their multiple relationships :

[0038] ;

[0039] Where k is the total number of key entities, are the weight coefficients of temporal relationship, spatial relationship, causal relationship and logical association, which are used to adjust the importance of different types of relationships in multiple evidence association analysis. The matrix elements represent temporal relationships, spatial relationships, causal relationships, and logical associations, respectively, and describe the specific relationships between different entities. is a binary function, defined as:

[0040] .

[0041] Optionally, S4 includes the following steps:

[0042] S41. From the relational dataset Extract each key entity And its related multiple relationships ,in, Represents key entities With another key entity Multiple evidence associations between them, including temporal relationships, spatial relationships, causal relationships, and logical relationships, for each key entity , which is related to another key entity The relationship between them is expressed as:

[0043] ;

[0044] in, , , and They are the influencing factors of temporal relationship, spatial relationship, causal relationship and logical association in multiple evidence association analysis;

[0045] S42. Constructing a weighted hypergraph structure of aggregated nodes ,in, is a set of aggregation nodes, each of which A collection of related key entities composition:

[0046] ;

[0047] Among them, f is the aggregation function, the entity set Aggregate into one node ;

[0048] S43. In the weighted hypergraph structure with aggregated nodes, each hyperedge Connect multiple aggregation nodes, combine the multiple associations between multiple entities within each aggregation node, and weight them for the calculation of hyperedge weights. Calculated by the following formula:

[0049] ;

[0050] in, , , and is the weight coefficient of each type of relationship in the hyperedge weight calculation, Indicates that on the edge The two aggregation nodes connected in and , Respectively represent the aggregation nodes and The specific entities contained in and It is an aggregation node and The entity set in and is a single entity in a collection of entities;

[0051] S44. Weighted hypergraph using aggregated nodes Model and represent the relationship between multiple entities. For each aggregation node and its connected hyperedges , the total relationship strength between aggregated nodes is expressed as:

[0052] ;

[0053] in, is the indicator function, when the aggregation node and Through the super edge The value is 1 when directly or indirectly connected, otherwise it is 0. Represents the total association strength between aggregated nodes in the hypergraph.

[0054] Optionally, S5 includes the following steps:

[0055] S51. Constructing the initial case record knowledge graph:

[0056] ;

[0057] in, Represents a node set. The initial node set is an empty set, which will be used to store entities extracted from case records. Represents the edge set. The initial edge set is an empty set, which will be used to store the relationship between entities.

[0058] S52, aggregate nodes into a weighted hypergraph Each aggregation node in Mapping to the case record knowledge graph Nodes in :

[0059] ;

[0060] in, is the mapping function, Represents an aggregation node The feature vectors of each entity in is the corresponding weight coefficient;

[0061] S53, aggregate nodes to the hyperedges in the weighted hypergraph Split into edges in the case record knowledge graph :

[0062] ;

[0063] in, is the hyperedge splitting function, is the relationship strength function, representing two aggregation nodes and The strength of the relationship between is the edge after splitting;

[0064] S54. Add the hyperedge weights in the aggregated node weighted hypergraph Mapped to the attribute weights of the case record knowledge graph edges :

[0065] ;

[0066] in, is the mapping function, is the weight mapping function, which transforms the weight information of the hyperedge Relationship after split Combined, is the weight coefficient, is the node’s impact factor, is the number of associated nodes;

[0067] S55. Introducing relationship nodes into the case record knowledge graph Or aggregation relationship:

[0068] ;

[0069] in, Build functions for relationship nodes, is the relationship weight coefficient, which indicates the importance of each node in the aggregation relationship. is the relationship influencing factor, is the weight calculation between the node and its associated relationship, is the adjustment factor of relationship strength.

[0070] Optionally, the S6 includes the following steps:

[0071] S61. Knowledge graph based on case records and aggregated node weighted hypergraph , define a graph query function to receive the query condition q and and Hypergraph Perform a joint query operation to generate a preliminary query result set :

[0072] ;

[0073] in, represents a node in the case knowledge graph, represents the edge in the graph, represents the aggregation node in the hypergraph, q is the query condition, which specifies the type and scope of multiple evidence associations to be retrieved;

[0074] S62. Query result set Conduct hypergraph analysis, apply hypergraph analysis functions to identify and deduce complex correlations between multiple pieces of evidence, and generate evidence correlation analysis results :

[0075] ;

[0076] in, is a hypergraph analysis function used to evaluate hyperedges Multiple evidence associations in For the indicator function, when the query result set With super edge The value is 1 when there is an intersection, otherwise it is 0;

[0077] S63. Results of evidence-based association analysis Derive potential correlations in the case and generate the final evidence correlation analysis report :

[0078] ;

[0079] in, is the association metric function between nodes and edges, is the mapping function between the edge association strength and the analysis result, Represents the final evidence association analysis report, which includes all relevant multiple evidence associations in the case and their derivation results.

[0080] Optionally, the S7 includes the following steps:

[0081] S71. Evidence-based correlation analysis report , extract key evidence associations, and define key evidence association sets , used to indicate the main evidence related to the core facts of the case and its relationship;

[0082] S72. According to the key evidence association set , derive the case progress path, define the progress path set , used to represent the order of events in a case and their associated paths:

[0083] ;

[0084] in, Indicates the case progress path. is a path generation function that generates a path based on the key evidence association set. is the path importance measurement function, measuring the path Importance in the progress of the case, is the path importance threshold, and the paths exceeding this threshold are included in the progress path set;

[0085] S73. Identify and mark potential evidence conflicts and define evidence conflict sets , used to record the contradictory or inconsistent evidence in the case and their relationship;

[0086] S74. Integrate the results of steps S71 to S73 to generate a final case analysis report, which includes key evidence associations, case progress paths, and potential conflicts of evidence.

[0087] The beneficial effects of the present invention are:

[0088] (1) The present invention pre-processes unstructured case records through natural language processing technology, including text segmentation, part-of-speech tagging, and syntactic analysis, automatically extracts key entities and relationships in the case, and maps them into the case record knowledge graph in a structured manner. This not only improves the accuracy of information extraction, but also lays a solid foundation for the subsequent intelligent association analysis of multiple evidences. Compared with the traditional single information extraction method, the present invention can efficiently and accurately identify key evidence and the relationships between them in large-scale case data, significantly improving the automation level and processing efficiency of case analysis.

[0089] (2) The present invention introduces an aggregated node weighted hypergraph structure. By aggregating related entities into nodes and modeling and analyzing the complex relationships between multiple entities in a weighted hypergraph, the relationship representation in complex cases is greatly simplified. Aggregating nodes not only reduces the complexity of the hypergraph, so that the graph structure remains efficient and clear in large-scale case data, but also can accurately represent the strength and importance of the relationship between different evidence through weighted processing of hyperedges. When dealing with complex cases involving multiple witnesses and physical evidence, it can more intuitively display the multi-dimensional relationship between evidence and reduce computational complexity, providing judicial personnel with a clearer basis for analysis.

[0090] (3) The present invention combines the case record knowledge graph with the weighted hypergraph of aggregated nodes to conduct an in-depth analysis of the complex correlation relationships between multiple pieces of evidence. It also automatically identifies and derives the potential correlation relationships in the case through graph query and hypergraph analysis technology, covering the correlations in multiple dimensions of time, space, cause and effect, and logic, generating complete evidence correlation analysis results and providing all-round decision support for case trials.

[0091] (4) Based on the results of multiple evidence association analysis, the present invention generates a case analysis report that includes key evidence associations, case progress paths, evidence relationship structures, and potential evidence conflicts. It intuitively displays the core evidence in the case and its association paths, and can mark possible evidence conflicts. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] The accompanying 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 of the present invention. In the accompanying drawings:

[0093] Figure 1 This is a flow chart of a method for associating multiple pieces of evidence in case records based on a knowledge graph proposed by the present invention;

[0094] Figure 2 This is a schematic diagram of the construction of a weighted hypergraph of aggregated nodes in a method for associating multiple evidences in case records based on a knowledge graph proposed in the present invention and its mapping in the knowledge graph. DETAILED DESCRIPTION

[0095] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0096] refer to Figure 1-2 , a method for associating multiple evidences in case records based on knowledge graph, comprising the following steps:

[0097] S1. Use natural language processing technology to pre-process unstructured case records, including word segmentation, part-of-speech tagging, and syntactic analysis of the text to generate a basic data set for semantic analysis;

[0098] S2. Perform named entity recognition on the basic data set to extract key entities in the case transcripts, including people, places, time, events, and evidence information, and mark the key entities as structured data;

[0099] S3, extract relationships from the extracted structured data to identify the temporal, spatial, causal relationships between different entities and the logical associations between evidence;

[0100] S4. Construct a weighted hypergraph structure and use the weighted hypergraph to model and represent the relationship between multiple entities. The weighted hypergraph nodes represent the key entities in the case records, the hyperedges represent the associations between different entities, and the weights of the hyperedges are used to describe the strength or importance of the associations.

[0101] S5, mapping the node and hyperedge information in the weighted hypergraph to the case record knowledge graph;

[0102] S6. Use graph query and hypergraph analysis technology to jointly analyze the constructed case record knowledge graph and weighted hypergraph, automatically identify and deduce the association relationship between multiple evidences in the case, and generate evidence association analysis results;

[0103] S7. Generate a case analysis report based on the results of the evidence correlation analysis.

[0104] In this implementation, S1 includes the following steps:

[0105] S11, performing word segmentation processing on the text data of the unstructured case record, splitting the continuous original text data T of the case record into word sequences W;

[0106] S12, perform part-of-speech tagging on the word sequence W to generate a part-of-speech tagging sequence P;

[0107] S13, perform syntactic analysis on the part-of-speech tag sequence P to generate a syntactic dependency tree D:

[0108] D = (N, E);

[0109] Where N represents the node set in syntactic analysis, each node corresponds to a word and its part-of-speech tag, E represents the edge set of syntactic dependency, and each edge represents the grammatical dependency between words;

[0110] S14. Extract the subject-verb-object structure and modification relationship in the sentence based on the syntactic dependency tree D to form a structured semantic fragment ;

[0111] S15. Semantic fragments Constructed as a basic dataset for semantic analysis:

[0112] ;

[0113] Among them, the basic data set Including semantic fragments , part-of-speech tag sequence P and word sequence W.

[0114] In this implementation, S2 includes the following steps:

[0115] S21. Basic data sets from semantic analysis Each semantic fragment Perform named entity recognition to identify the key entity E in the case record. The key entities include people, places, time, events, and evidence information, and generate an entity set:

[0116] ;

[0117] in, represents a single key entity in the named entity recognition process, and k is the number of key entities identified;

[0118] S22. For each key entity Mark the attribute as structured data ;

[0119] S23, the marked structured data With the original basic data set Perform association to generate an extended structured dataset :

[0120] .

[0121] In this implementation, S3 includes the following steps:

[0122] S31. From the extended structured dataset Extract each key entity and associated attribute information ;

[0123] S32. Extracting key entities Analyze the associations with other entities to identify the temporal relationships between entities , time relationship It is defined as the order of entities on the time axis, expressed as;

[0124] S33. Identify spatial relationships between entities , where the spatial relationship Defined as the relative relationship of entities in space;

[0125] S34. Identify causal relationships between entities ,causation Defined as a key entity The occurrence of Occurrence of;

[0126] S35. Identify logical associations between entities , logical association Defined as the existence of interdependent or contradictory relationships between pieces of evidence;

[0127] S36: Combine the time relationship, spatial relationship, causal relationship and logical relationship identified in steps S32 to S35 to conduct a comprehensive correlation analysis of multiple pieces of evidence, identify the relationship between multiple pieces of evidence in the case record, and integrate them into the extended structured data set. Generate a dataset containing entities and their multiple relationships :

[0128] ;

[0129] Where k is the total number of key entities, are the weight coefficients of temporal relationship, spatial relationship, causal relationship and logical association, which are used to adjust the importance of different types of relationships in multiple evidence association analysis. The matrix elements represent temporal relationships, spatial relationships, causal relationships, and logical associations, respectively, and describe the specific relationships between different entities. is a binary function, defined as:

[0130] .

[0131] In this implementation, S4 includes the following steps:

[0132] S41. From the relational dataset Extract each key entity And its related multiple relationships ,in, Represents key entities With another key entity Multiple evidence associations between them, including temporal relationships, spatial relationships, causal relationships, and logical relationships, for each key entity , which is related to another key entity The relationship between them is expressed as:

[0133] ;

[0134] in, , , and They are the influencing factors of temporal relationship, spatial relationship, causal relationship and logical association in multiple evidence association analysis;

[0135] S42. Constructing a weighted hypergraph structure of aggregated nodes ,in, is a set of aggregation nodes, each of which A collection of related key entities composition:

[0136] ;

[0137] Among them, f is the aggregation function, the entity set Aggregate into one node ;

[0138] S43. In the weighted hypergraph structure with aggregated nodes, each hyperedge Connect multiple aggregation nodes, combine the multiple associations between multiple entities within each aggregation node, and weight them for the calculation of hyperedge weights. Calculated by the following formula:

[0139] ;

[0140] in, , , and is the weight coefficient of each type of relationship in the hyperedge weight calculation, Indicates that on the edge The two aggregation nodes connected in and , Respectively represent the aggregation nodes and The specific entities contained in and It is an aggregation node and The entity set in and is a single entity in a collection of entities;

[0141] S44. Weighted hypergraph using aggregated nodes Model and represent the relationship between multiple entities. For each aggregation node and its connected hyperedges , the total relationship strength between aggregated nodes is expressed as:

[0142] ;

[0143] in, is the indicator function, when the aggregation node and Through the super edge The value is 1 when directly or indirectly connected, otherwise it is 0. Represents the total association strength between aggregated nodes in the hypergraph.

[0144] In this implementation, S5 includes the following steps:

[0145] S51. Constructing the initial case record knowledge graph:

[0146] ;

[0147] in, Represents a node set. The initial node set is an empty set, which will be used to store entities extracted from case records. Represents the edge set. The initial edge set is an empty set, which will be used to store the relationship between entities.

[0148] S52, aggregate nodes into a weighted hypergraph Each aggregation node in Mapping to the case record knowledge graph Nodes in :

[0149] ;

[0150] in, is the mapping function, Represents an aggregation node The feature vectors of each entity in is the corresponding weight coefficient;

[0151] S53, aggregate nodes to the hyperedges in the weighted hypergraph Split into edges in the case record knowledge graph :

[0152] ;

[0153] in, is the hyperedge splitting function, is the relationship strength function, representing two aggregation nodes and The strength of the relationship between is the edge after splitting;

[0154] S54. Add the hyperedge weights in the aggregated node weighted hypergraph Mapped to the attribute weights of the case record knowledge graph edges :

[0155] ;

[0156] in, is the mapping function, is the weight mapping function, which transforms the weight information of the hyperedge Relationship after split Combined, is the weight coefficient, is the node’s impact factor, is the number of associated nodes;

[0157] S55. Introducing relationship nodes into the case record knowledge graph Or aggregation relationship:

[0158] ;

[0159] in, Build functions for relationship nodes, is the relationship weight coefficient, which indicates the importance of each node in the aggregation relationship. is the relationship influencing factor, is the weight calculation between the node and its associated relationship, is the adjustment factor of relationship strength.

[0160] In this implementation, S6 includes the following steps:

[0161] S61. Knowledge graph based on case records and aggregated node weighted hypergraph , define a graph query function to receive the query condition q and and Hypergraph Perform a joint query operation to generate a preliminary query result set :

[0162] ;

[0163] in, represents a node in the case knowledge graph, represents the edge in the graph, represents the aggregation node in the hypergraph, q is the query condition, which specifies the type and scope of multiple evidence associations to be retrieved;

[0164] S62. Query result set Conduct hypergraph analysis, apply hypergraph analysis functions to identify and deduce complex correlations between multiple pieces of evidence, and generate evidence correlation analysis results :

[0165] ;

[0166] in, is a hypergraph analysis function used to evaluate hyperedges Multiple evidence associations in For the indicator function, when the query result set With super edge The value is 1 when there is an intersection, otherwise it is 0;

[0167] S63. Results of evidence-based association analysis Derive potential correlations in the case and generate the final evidence correlation analysis report :

[0168] ;

[0169] in, is the association metric function between nodes and edges, is the mapping function between the edge association strength and the analysis result, Represents the final evidence association analysis report, which includes all relevant multiple evidence associations in the case and their derivation results.

[0170] In this implementation, S7 includes the following steps:

[0171] S71. Evidence-based correlation analysis report , extract key evidence associations, and define key evidence association sets , used to indicate the main evidence related to the core facts of the case and its relationship;

[0172] S72. According to the key evidence association set , derive the case progress path, define the progress path set , used to represent the order of events in a case and their associated paths:

[0173] ;

[0174] in, Indicates the case progress path. is a path generation function that generates a path based on the key evidence association set. is the path importance measurement function, measuring the path Importance in the progress of the case, is the path importance threshold, and the paths exceeding this threshold are included in the progress path set;

[0175] S73. Identify and mark potential evidence conflicts and define evidence conflict sets , used to record contradictory or inconsistent evidence in a case and their relationship:

[0176] ;

[0177] in, is the conflict measurement function, measuring the evidence and evidence The degree of contradiction between the evidence is the key entity, is the conflict degree threshold, and the evidence pairs exceeding this threshold are considered to be in conflict and included in the evidence conflict set;

[0178] S74. Integrate the results of steps S71 to S73 to generate a final case analysis report, which includes key evidence associations, case progress paths, and potential conflicts of evidence.

[0179] The present invention pre-processes unstructured case transcripts through natural language processing technology, including text segmentation, part-of-speech tagging, and syntactic analysis, automatically extracts key entities and relationships in the case, and maps them in a structured manner to the case transcript knowledge graph, which not only improves the accuracy of information extraction, but also lays a solid foundation for subsequent intelligent association analysis of multiple evidences. Compared with traditional single information extraction methods, the present invention can efficiently and accurately identify key evidence and the relationships between them in large-scale case data, significantly improving the degree of automation and processing efficiency of case analysis.

[0180] The present invention introduces an aggregated node weighted hypergraph structure, which greatly simplifies the relationship representation in complex cases by aggregating related entities into nodes and modeling and analyzing the complex relationships between multiple entities in a weighted hypergraph. Aggregated nodes not only reduce the complexity of the hypergraph, so that the graph structure remains efficient and clear in large-scale case data, but also can accurately represent the strength and importance of the association between different evidence through weighted processing of hyperedges. When dealing with complex cases involving multiple witnesses and physical evidence, it can more intuitively display the multi-dimensional association between evidence and reduce computational complexity, providing judicial personnel with a clearer basis for analysis.

[0181] The present invention combines the case record knowledge graph with the aggregated node weighted hypergraph to conduct an in-depth analysis of the complex correlation relationships between multiple pieces of evidence, and automatically identifies and derives potential correlation relationships in the case through graph query and hypergraph analysis technology, covering multi-dimensional correlations in time, space, causality and logic, generating complete evidence correlation analysis results, and providing all-round decision-making support for case trials.

[0182] Based on the results of multiple evidence association analysis, the present invention generates a case analysis report that includes key evidence associations, case progress paths, evidence relationship structures, and potential evidence conflicts. It intuitively displays the core evidence in the case and its association paths, and can mark possible evidence conflicts.

[0183] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for associating multiple evidences in case records based on knowledge graph, characterized in that: The steps include: S1. Use natural language processing technology to pre-process unstructured case records, including word segmentation, part-of-speech tagging, and syntactic analysis of the text to generate a basic data set for semantic analysis; S2. Perform named entity recognition on the basic data set to extract key entities in the case transcripts, including people, places, time, events, and evidence information, and mark the key entities as structured data; S3, extract relationships from the extracted structured data to identify the temporal, spatial, causal relationships between different entities and the logical associations between evidence; S4. Construct a weighted hypergraph structure, and use the weighted hypergraph to model and represent the relationship between multiple entities, wherein the weighted hypergraph nodes represent key entities in the case record, the hyperedges represent the association relationship between different entities, and the weights of the hyperedges are used to describe the strength or importance of the association relationship; S5, mapping the node and hyperedge information in the weighted hypergraph to the case record knowledge graph; S6. Use graph query and hypergraph analysis technology to jointly analyze the constructed case record knowledge graph and weighted hypergraph, automatically identify and deduce the association relationship between multiple evidences in the case, and generate evidence association analysis results; S7. Generate a case analysis report based on the evidence correlation analysis results; The S4 comprises the following steps: S41. From the relational dataset D r Extract each key entity e i and its related multiple associations R multi (e i ,e j ), where R multi (e i ,e j ) represents the key entity e i With another key entity e j The multiple evidence association relationships between them include the combination of temporal relationship, spatial relationship, causal relationship and logical association. For each key entity e i , which is related to another key entity e j The relationship between them is expressed as: R multi (e i ,e j )6γ t ·R t (e i ,e j )+γ s ·R s (e i ,e j )+γ c ·R c (e i ,e j )+γ l ·R l (e i ,e j )4 Among them, γ t , γ s , γ c and γ l They are the influencing factors of temporal relationship, spatial relationship, causal relationship and logical association in multiple evidence association analysis; S42. Constructing a weighted hypergraph structure H of aggregated nodes g =(V g ,E h ,W h ), where V g =v1,v2,...,v m } is a set of aggregation nodes, each aggregation node v i A collection of related key entities composition: Among them, f is the aggregation function, the entity set Aggregate into a node v i ; S43. In the weighted hypergraph structure with aggregated nodes, each hyperedge Connect multiple aggregation nodes, combine the multiple associations between multiple entities in each aggregation node, and weight them for the calculation of hyperedge weights. The hyperedge weight w i Calculated by the following formula: Among them, β t , β s , β c and β l is the weight coefficient of each type of relationship in the hyperedge weight calculation, Indicates that on the edge The two aggregation nodes v connected in x and v y , Respectively represent the aggregation node v x and v y The specific entities contained in and is the aggregation node v x and v y The entity set in e xa and e yb is a single entity in a collection of entities; S44. Using the weighted hypergraph H using aggregated nodes g Model and represent the relationship between multiple entities. For each aggregation node v i and its connected hyperedges The total relationship strength between aggregated nodes is expressed as: in, is the indicator function, when the aggregation node v i and v j Through the super edge The value is 1 when directly or indirectly connected, otherwise it is 0. total (v i ,v j ) represents the total association strength between aggregated nodes in the hypergraph; The S5 comprises the following steps: S51. Constructing the initial case record knowledge graph: G k =(N k ,R k ); Among them, N k Represents a node set. The initial node set is an empty set, which will be used to store entities extracted from case records. k Represents the edge set. The initial edge set is an empty set, which will be used to store the relationship between entities. S52, aggregate the nodes into a weighted hypergraph H g Each aggregation node v in i Mapped to the case record knowledge graph G k Node n in i : Among them, map(v i ) is the mapping function, e ij Represents the aggregation node v i The feature vector of each entity in , λ j is the corresponding weight coefficient; S53, aggregate nodes to the hyperedges in the weighted hypergraph Split into edges r in the case record knowledge graph ij : in, is the hyperedge splitting function, φ(v x ,v y ) is the relationship strength function, representing two aggregation nodes v x and v y The strength of the relationship between xy is the edge after splitting; S54, adding the hyperedge weight w in the weighted hypergraph of the aggregated node i Mapped to the attribute weights of the case record knowledge graph edges Among them, f(w i ) is the mapping function, ψ(w i ,r xy ) is the weight mapping function, which transforms the weight information w of the hyperedge i Relationship with split xy Combined, α xy is the weight coefficient, θ k is the node’s influence factor, n k is the number of associated nodes; S55. Introducing relationship node n in the case record knowledge graph r Or aggregation relationship: Among them, g is the relationship node construction function, μ i is the relationship weight coefficient, indicating the importance of each node in the aggregation relationship, ν j is the relationship impact factor, φ(n ij ,r j ) is the weight calculation between the node and its association relationship, κ k is the relationship strength moderator; The S6 comprises the following steps: S61. Knowledge graph based on case records G k and the aggregated node weighted hypergraph H g , define a graph query function to receive the query condition q and k and Hypergraph H g Perform the joint query operation in the query to generate the preliminary query result set R q : Among them, n i Represents a node in the case knowledge graph, r ij represents the edge in the graph, v x represents the aggregation node in the hypergraph, q is the query condition, which specifies the type and scope of multiple evidence associations to be retrieved; S62, query result set R q Perform hypergraph analysis, apply hypergraph analysis functions to identify and deduce complex correlations between multiple pieces of evidence, and generate evidence correlation analysis results R a : in, is a hypergraph analysis function used to evaluate hyperedges Multiple evidence associations in is the indicator function. When the query result set R q With super edge The value is 1 when there is an intersection, otherwise it is 0; S63. Based on the results of evidence association analysis R a Derive the potential correlation relationship in the case and generate the final evidence correlation analysis report R f .

2. According to claim 1, a method for associating multiple evidences in case records based on knowledge graph, characterized in that: The S1 comprises the following steps: S11, performing word segmentation processing on the text data of the unstructured case record, splitting the continuous original text data T of the case record into word sequences W; S12, perform part-of-speech tagging on the word sequence W to generate a part-of-speech tagging sequence P; S13, perform syntactic analysis on the part-of-speech tag sequence P to generate a syntactic dependency tree D: D = (N, E); Where N represents the node set in syntactic analysis, each node corresponds to a word and its part-of-speech tag, E represents the edge set of syntactic dependency, and each edge represents the grammatical dependency between words; S14. Extract the subject-verb-object structure and modification relationship in the sentence based on the syntactic dependency tree D to form a structured semantic fragment S f ; S15, the semantic fragment S f Constructed as a basic dataset for semantic analysis: D b ={S f ,P,W}; Among them, the basic data set D b Including semantic fragment S f , part-of-speech tag sequence P and word sequence W.

3. According to claim 1, a method for associating multiple evidences in case records based on knowledge graph, characterized in that: The S2 comprises the following steps: S21. Basic dataset D from semantic analysis b Each semantic segment S f Perform named entity recognition to identify key entities E in the case transcript, including person, place, time, event, and evidence information, and generate an entity set: E}{e1,e2,...,e k }; Among them, e i represents a single key entity in the named entity recognition process, and k is the number of key entities identified; S22. For each key entity e i Mark the attributes as structured data S d ; S23, the labeled structured data S d Compared with the original basic data set D b Perform association to generate an extended structured dataset D s : D s ={S d ,S f ,P,W}。 4. According to claim 1, a method for associating multiple evidences in case records based on knowledge graph, characterized in that: The S3 comprises the following steps: S31. From the extended structured dataset D s Extract each key entity e i and associated attribute information A i ; S32, extract the key entity e i Analyze the association with other entities and identify the temporal relationship between entities R t , time relationship R t It is defined as the order of entities on the time axis, expressed as; S33. Identify the spatial relationship between entities R s , where the spatial relationship R s Defined as the relative relationship of entities in space; S34. Identify causal relationships between entities R c , causal relationship R c Defined as a key entity e i The occurrence of causes or affects another entity j Occurrence of; S35. Identify logical associations between entities R l , logical association R l Defined as the existence of interdependent or contradictory relationships between pieces of evidence; S36: Combine the time relationship, spatial relationship, causal relationship and logical relationship identified in steps S32 to S35 to conduct a comprehensive correlation analysis of multiple pieces of evidence, identify the relationship between multiple pieces of evidence in the case record, and integrate them into the extended structured data set D s In this paper, a dataset D containing entities and their multiple associations is generated. r : Where k is the total number of key entities, α t , α s , α c , α l are the weight coefficients of temporal relationship, spatial relationship, causal relationship and logical association, which are used to adjust the importance of different types of relationships in multiple evidence association analysis. t (e i ,e j ), R s (e i ,e j ), R c (e i ,e j ), R l (e i ,e j ) represent the matrix elements of temporal relationship, spatial relationship, causal relationship and logical association, respectively, describing the specific relationship between different entities, δ(e i ,e j ) is a binary function, defined as:

5. According to claim 1, a method for associating multiple evidences in case records based on knowledge graph, characterized in that: The S7 comprises the following steps: S71. Report R based on evidence association analysis f The analysis results in extract key evidence associations and define the key evidence association set C e , used to indicate the main evidence related to the core facts of the case and its relationship; S72. According to the key evidence association set C e , derive the case progress path, define the progress path set P a , used to represent the order of events in a case and their associated paths: Among them, p j represents the case progress path, σ(C e ) is the path generation function, which generates the path according to the key evidence association set, α p (p j ) is the path importance measurement function, which measures the path p j Importance in case development, τ p is the path importance threshold, and the paths exceeding this threshold are included in the progress path set; S73. Identify and mark potential evidence conflicts and define the evidence conflict set C f , used to record the contradictory or inconsistent evidence in the case and their relationship; S74. Integrate the results of steps S71 to S73 to generate a final case analysis report, which includes key evidence associations, case progress paths, and potential conflicts of evidence.

Images