A legal document data sharing method based on data analysis

Abstract

The application relates to the technical field of data processing, in particular to a legal document data sharing method based on data analysis, which comprises the following steps: cutting a legal document text according to a part of speech boundary, dividing the legal document into multiple sharing nodes according to legal element parts of speech, performing structural analysis on each sharing node, extracting structural data corresponding to each sharing node based on the integrated relationship of each piece of information in the sharing node, performing interactive analysis on the structural data of each sharing node, outputting the analysis result of the current legal document in the form of clustering of a similar set and a conflict set after multiple text inputs in a parameter interaction mode, performing abnormal clustering detection based on the analysis result of the current legal document, identifying difference nodes between the sharing nodes in the access order of the sharing nodes, connecting the identified difference nodes with adjacent sharing nodes, and completing the configuration of each sharing node. The method improves the efficiency of legal document processing and the consistency of structural division.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of data processing, and in particular to a legal document data sharing method based on data analysis. BACKGROUND

[0002] Legal documents are mainly described in natural language and include judgment books, ruling books, mediation books and decision books. Due to the great differences in the formats and terminology systems of legal documents of different institutions such as courts, law firms and enterprises, it is difficult to identify the text description content, resulting in difficulties in legal document management.

[0003] For example, Chinese Patent Publication No. CN119149569A discloses a legal supervision clue mining method and system based on full-text retrieval and large models, which belongs to the field of data collection and data retrieval and large model application. The method comprises the following steps: S1. Combining supervision rules and retrieval keywords; S2. Full-text retrieval, using the keywords and full-text retrieval system to perform data retrieval, collecting all related graphic and text data within a specified time period from multiple data sources; S3. Information distillation and information filtering; S4. Building a knowledge base for topic analysis; S5. Knowledge base retrieval and sorting; S6. Large model intelligent analysis and clue mining.

[0004] For example, Chinese Patent Publication No. CN113239206A discloses a method for precise data classification and analysis of judicial documents and a computer-readable storage device, which comprises a storage device and the following steps: A) establishing a legal relationship classification word library; B) identifying electronic tags; C) field extraction and electronic tag classification of judicial documents; D) user selective correction, feeding the correction logic back to the legal relationship classification word library; E) outputting user-defined charts directly related to the data or directly related to the results of multiple linear analysis.

[0005] The prior art describes a comprehensive retrieval of keywords and related images to complete the retrieval and processing of legal documents, and uses data type tags of judicial documents to complete the processing of legal documents in the form of chart display. However, in the entire processing of legal documents, it is necessary to identify the format content of legal documents from different sources and the difference nodes existing in different contents under the format, and to complete the sharing processing of legal documents by consistency checking of the difference nodes. SUMMARY

[0006] To solve the above technical problems, the technical solution adopted by the application is: a legal document data sharing method based on data analysis, comprising: S1, cutting the legal document text according to the part of speech boundary, dividing the legal document into multiple sharing nodes according to the legal element part of speech, and recording the embedding dimension of each sharing node.

[0007] S2 performs structural analysis on each shared node, and based on the integration relationship of each piece of information within the shared node, performs data tracing on the shared node and extracts the structural data corresponding to each shared node.

[0008] S3 performs interactive analysis on the structural data corresponding to each shared node, obtains the parameter interaction patterns between each shared node, and outputs the analysis results of the current legal document in the form of clustering similar sets and conflict sets after multiple text inputs based on the parameter interaction patterns.

[0009] S4, based on the analysis results of the current legal documents, performs anomaly clustering detection on the input legal documents, and identifies the difference nodes between the shared nodes according to the access order of each shared node.

[0010] S5 connects the identified difference nodes with adjacent shared nodes, sets up node transaction processes for shared nodes, and completes the configuration of each shared node.

[0011] The beneficial effects of this invention are as follows: First, by using part-of-speech boundaries, this invention divides legal document text into multiple shared nodes based on the part-of-speech attributes of multiple legal elements. Then, based on the position, sub-words, tags, and sentences of the segmented text, each shared node is displayed in a structured form, thereby realizing the data combination and segmentation of legal documents under different scenarios.

[0012] Second, this invention integrates legal documents from different sources through a hierarchical structure and the association between supplementary legal documents, and achieves traceable integration processing of legal documents by updating the content of corresponding nodes at different times. Then, by using the legal facts in the legal documents, multiple sets of legal documents are formed into similar sets and conflict sets in a progressive and overlapping manner, so as to achieve overall identification and processing of conflicts and similarities between legal documents from multiple sources, thereby further improving the collaborative sharing method of multiple legal documents.

[0013] Third, this invention records the access order of shared nodes, identifies the relative matching threshold of the selected difference nodes, and then merges and analyzes the difference nodes with the shared nodes, binding them in the form of key-value pairs. This illustrates the key-value pair situation when legal documents are shared under multiple sources and different formats. By using the time interval when nodes are updated, the order of difference nodes and adjacent nodes during node transaction updates is determined, and the corresponding shared nodes are displayed one by one. This enables a comparison and display of the speed of shared data updates and the content of shared data, improving the accuracy and interpretability of shared nodes. Attached Figure Description

[0014] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0015] Figure 1This is a flowchart illustrating a data-driven approach to sharing legal document data.

[0016] Figure 2 This is a schematic diagram of a data sharing method for legal documents based on data analysis.

[0017] Figure 3 This is a flowchart illustrating step S1 of a data analysis-based legal document data sharing method.

[0018] Figure 4 This is a flowchart illustrating step S2 of a data analysis-based legal document data sharing method.

[0019] Figure 5 This is a flowchart illustrating step S3 of a data analysis-based legal document data sharing method.

[0020] Figure 6 This is a schematic diagram of step S3 of a data analysis-based legal document data sharing method.

[0021] Figure 7 This is a flowchart illustrating step S4 of a data analysis-based legal document data sharing method. Detailed Implementation

[0022] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in the art or in accordance with the product manual.

[0023] See Figure 1 , Figure 2 A data analysis-based method for sharing legal documents includes: S1, segmenting legal document texts according to part-of-speech boundaries, dividing legal documents into multiple shared nodes according to the part-of-speech of legal elements, and recording the embedding dimension of each shared node.

[0024] S2 performs structural analysis on each shared node, and based on the integration relationship of each piece of information within the shared node, performs data tracing on the shared node and extracts the structural data corresponding to each shared node.

[0025] S3 performs interactive analysis on the structural data corresponding to each shared node, obtains the parameter interaction patterns between each shared node, and outputs the analysis results of the current legal document in the form of clustering similar sets and conflict sets after multiple text inputs based on the parameter interaction patterns.

[0026] S4, based on the analysis results of the current legal documents, performs anomaly clustering detection on the input legal documents, and identifies the difference nodes between the shared nodes according to the access order of each shared node.

[0027] S5 connects the identified difference nodes with adjacent shared nodes, sets up node transaction processes for shared nodes, and completes the configuration of each shared node.

[0028] Legal documents contain a significant amount of personal information, financial details, and criminal records. They are composed of highly structured and fixed formats, with complex and multi-dimensional structures. Therefore, it is necessary to use anonymized data to create a relationship graph that corresponds to the parties, lawyers, courts, and legal provisions. This graph should be used to establish connections based on legal provisions and to push legal documents through attribution or association, thus enabling the shared analysis of legal documents.

[0029] When analyzing and setting up shared nodes, the shared nodes are divided based on part-of-speech dependency, part-of-speech constraint, and part-of-speech boundary information. Each shared node is set based on the sub-word embedding, sentence embedding, position embedding, and tag embedding of the data corresponding to the shared node. The embedding dimension of the shared node is used as the main target object of the current sharing.

[0030] like Figure 3 As shown, the implementation of step S1 includes: S11, when a requester accesses a legal document, semantic analysis is performed on the current legal document, and the legal element parts of speech are set using preset part-of-speech boundaries as word segmentation conditions. When segmenting legal element parts of speech, based on the completeness of the descriptions of various legal elements, multiple words are combined into legal element parts of speech through semantic analysis to represent the parties involved, criminal acts, legal citations, chains of evidence, etc., in the current legal document, and their related content is treated as a set of related data for legal element parts of speech.

[0031] Preferably, the legal elements described above are categorized into multiple shared nodes based on the meaning of the words they contain, such as parties = noun phrases, criminal acts = verb structures, legal citations = numerical codes + clauses, evidence chains = multiple nouns connected by coordinating conjunctions, and judgment results = legal actions + modal verbs. The part-of-speech boundaries are used to describe the vocabulary representations, nouns, verbs, objects, clauses, numerical descriptions, etc., within the legal document, thus segmenting it into multiple shared nodes. The position of each shared node in the text before segmentation is recorded as a position embedding, and the tags of the shared nodes before segmentation are recorded as tag embeddings. Tag embeddings can be of various types, such as parties, legal provisions, and evidence, forming a node network. Simultaneously, the embedded shared nodes indicate whether each node has superior and subordinate descriptions, distinguishing whether the current shared node is a sub-word or summary term of other words.

[0032] S12 uses the data corresponding to the parts of speech of the legal elements in the current legal document as initial nodes. It defines the subordinate components of each initial node within the part-of-speech boundary, identifies the subordinate sub-words of each initial node, and sequentially performs position embedding, sub-word embedding, and sentence embedding based on the sentence and position of the subordinate sub-words, thus completing the boundary expansion of the initial nodes. At this point, boundary identification is performed on the divided initial nodes to find the parts of the current legal text that are continuously cited regarding criminal acts, legal citations, chains of evidence, etc. Then, the content described in these parts within the same paragraph or sentence is added to the corresponding initial nodes to complete the boundary expansion.

[0033] S13, use the labels of each initial node to perform label embedding, regard the initial nodes after embedding as the output shared nodes, and use the embedded data as the embedding dimension of each shared node.

[0034] Preferably, the above-mentioned position embedding means that the position of the current initial node and the specific position of the sub-word within the paragraph are associated with the corresponding shared node by the paragraph number, sentence position and character offset, and the character offset indicates the specific position number in the sentence.

[0035] Preferably, the aforementioned tags will include, but are not limited to, the parties, institutions, plaintiffs, and defendants corresponding to entity tags, criminal acts and legal procedures corresponding to behavioral tags, and procedural law and substantive law corresponding to legal provision tags, and these tags will be displayed in combination with the content after embedding sub-words.

[0036] Preferably, the above-mentioned sub-word embedding is accomplished by combining multiple words that have a subordinate relationship.

[0037] Preferably, the above sentence embedding involves summarizing the meaning of the corresponding sentence using legal rhetoric and combining the sub-words and the sentence extracted from the current initial node, and then adding them to the corresponding shared node.

[0038] In one embodiment of the present invention, by using legal documents as structural analysis, the analyzed shared nodes are determined. Based on the integration relationship constructed by each piece of information, the shared nodes are set up with multiple paths. Through multi-data tracing, the structural data corresponding to each shared node in the legal documents is set up, and the tracing analysis process of structural data on the shared node path is explained; to query the similar parts and abnormal conflict parts existing in each shared node.

[0039] When identifying shared nodes, it is also necessary to check whether the shared nodes can be connected, such as whether there are multiple forms of association, such as vertical association and horizontal association, and then connect these extracted shared nodes.

[0040] Vertical nodes indicate the existence of superior and subordinate nodes in the current shared node, horizontal associations indicate the existence of strongly related nodes in the current shared node, and the path representing the chronological evolution in the current legal document, such as describing the chronological progression of criminal acts, are used to determine the content contained in each shared node.

[0041] When describing the structural analysis of shared nodes, they can be integrated using four structures: legal structure, factual structure, evidentiary structure, and temporal structure. The integration relationships include vertical integration, horizontal integration, and dynamic integration. This completes the establishment of the source analysis network for the current shared nodes, and based on the output of the source network, the structural data corresponding to each shared node is obtained.

[0042] Vertical integration refers to the hierarchical relationship of nodes in legal concepts, i.e., a superior-subordinate relationship; horizontal integration refers to the logical connection of nodes in the same legal fact, such as complementary elements / supporting evidence / causal chains, etc., which are composed of multiple data in the same legal document; dynamic integration refers to the derivative connection of nodes in judicial practice, such as similar case adjudication rules / legal provisions evolution / regional differences.

[0043] The legal structure represents the legal system and usage relationships associated with the nodes; the factual structure represents the completeness of the legal fact description of the nodes; the evidence structure represents the main types of evidence that make up the current node; and the temporal structure represents the logical relationship between the shared nodes in time, to explain whether there is a relationship between the shared nodes.

[0044] like Figure 4 As shown, the implementation of step S2 also includes: S21, based on the legal structure, factual structure, evidence structure and temporal structure corresponding to the current shared node, determining whether the current shared node belongs to the vertical integration, horizontal integration and dynamic integration integration mode.

[0045] As for the integration method of shared nodes in step S21, the implementation method is as follows: S211, if the current shared node belongs to vertical integration, then based on the multiple sets of data of the shared node in vertical integration, the shared nodes corresponding to the vertical integration are connected in a hierarchical form to form a vertical hierarchical structure.

[0046] S212, if the current shared node belongs to horizontal integration, then a horizontal integration structure is formed based on the logical complementarity between multiple legal texts of the application.

[0047] S213, If the current shared nodes are in dynamic integration, a dynamic integration structure is formed based on the time sequence advancement between shared nodes under the same integration method.

[0048] Preferably, when forming a vertical hierarchical structure, the method further includes: establishing a hierarchical index for each shared node based on the embedding dimension of each shared node. This hierarchical index represents the contextual messages contained when each shared node performs word embedding, sentence embedding, and position embedding, and uses it as the level of each shared node in the vertical integration. The shared nodes are then integrated based on the hierarchical index of each shared node.

[0049] Preferably, when forming a horizontal integration structure, the method further includes: extracting sentence embedding and tag embedding data from the embedding dimension of each shared node; performing association rule mining on the sentence embedding and tag embedding data of the current shared node; and horizontally integrating shared nodes under the same association rule. Association rules are used to explain hidden relationships between different data, such as the association between crime patterns and geographical locations. At this point, keyword combinations are extracted from the data obtained from tag embedding and sentence embedding, and the shared nodes corresponding to different keyword combinations are horizontally integrated based on the frequency of the keyword combinations across multiple shared nodes. The nodes are then sorted from highest to lowest frequency based on the keyword combination.

[0050] Preferably, when forming a dynamic integration structure, the method further includes: checking the update time of the embedding dimension of the current shared node to update the changes in the vertical hierarchical structure and horizontal integration structure of the shared nodes before and after, as the dynamic integration structure. After the corresponding integration method is formed, the formed integration relationship is output as a chain relationship structure.

[0051] The final data integration will represent various chain relationship structures. For example, the factual element chain can include behavior → subjective element → objective result → causal relationship, and the adjudication rule chain can include the current adjudication → similar case rules → judicial policy → social effect, which will explain the structural information corresponding to the current legal document in turn. Then, according to the obtained integration relationship, it is necessary to explain the analysis results of data tracing under different integration relationships to extract the structural data that each shared node needs to output.

[0052] The implementation of step S2 also includes: S22, generating a structural fingerprint based on the current integration method of the shared nodes. The structural fingerprint is used to hash the structured path, generating a fixed-length binary fingerprint, and then displaying the relevant data of different shared nodes in tabular form. Path backtracking is performed based on the structural fingerprints of each shared node to identify the source analysis network corresponding to each shared node.

[0053] S23, explore all associated paths of the current shared node in the source analysis network, and output the associated paths of the current shared node as the structural data of each shared node.

[0054] In step S22 above, the part-of-speech boundaries under each integration method are used to control each integration method. Each shared node is represented using dots, and then the shared nodes are connected in an integrated manner to obtain the relevant paths of the shared nodes. Semantic annotations are then applied to the connections of shared nodes under each path, forming a chain of rules for chain-like connections, such as factual element chains, adjudication rule chains, and other different connection forms. The part-of-speech boundaries are used to represent the semantics of shared nodes when connected into paths, illustrating the information annotated after the connection.

[0055] Then, all paths related to the current shared node are output sequentially in a chain-like format to illustrate the structural data related to the current shared node.

[0056] The implementation of step S22, identifying the source analysis network corresponding to each shared node, includes: obtaining the connection method of shared nodes under different integration relationships, and identifying the element type based on multiple sets of shared nodes corresponding to the connection method. The element type includes the content of commonly used phrases in legal documents such as having children after marriage, premarital personal property, and joint property of husband and wife. These contents have a strongly correlated description, which facilitates the source tracing and verification of the connection relationship of legal documents in multiple contents.

[0057] Obtain the feature types of shared nodes under the corresponding group, and set the source tracing paths between shared nodes based on the pointing relationship of feature types in shared nodes. Combine the source tracing paths to form the source tracing analysis network.

[0058] If there are multiple intersecting parts between the tracing paths, the shared nodes of the intersecting parts are integrated based on their pointing relationships and referencing methods, and the integrated shared nodes point to other shared nodes.

[0059] The aforementioned directional relationships refer to the connection directions established between shared nodes through logical dependencies or semantic associations. For example: Node A (children after marriage) → Node B (child custody dispute), indicating that the existence of children is a prerequisite for a custody dispute; Node C (premarital personal property) → Node D (property division), indicating that premarital property needs to be handled separately in the division. The directional relationships between different shared nodes are categorized into causal, subordinate, referential, and parallel relationships to illustrate the combined relationships between multiple events in a case described in a legal document. The referential methods describe direct, indirect, and conditional references, thereby integrating multiple intersecting shared nodes.

[0060] In one embodiment of the present invention, the structural data is interactively analyzed. Under the classification scenarios of various types such as the legal provisions and parties involved in each data, similar texts are analyzed. The relationship condition process between the shared nodes is analyzed, and the relationship condition process is used as its parameter interaction mode. The interaction result of the parameter interaction mode in multiple text inputs is used as the analysis result of the current legal document.

[0061] In step S3, it is mainly necessary to use the parameter interaction of multiple acquired texts to deduce the similarity conflict detection of each node from the single relationship of shared nodes and node attributes, identify the content with partial conflicts, and the main content and shared nodes involved in the conflict when multiple legal documents are input. Based on these involved shared nodes, the analysis results of the current legal document under its data interaction and group decision evaluation are analyzed.

[0062] like Figure 5 , Figure 6 As shown, the implementation of step S3 includes: S31, viewing the legal facts of the shared nodes in the structural data, performing interactive analysis on each shared node based on the legal facts, and obtaining the progressive and overlapping expression methods in scenarios with similar word meanings.

[0063] The progressive presentation method refers to the gradual unfolding and deepening of a legal fact or element in different documents. For example, premarital property may only be mentioned in the lawsuit, the amount may be specified in the mediation agreement, and the ownership may be finally confirmed in the judgment. In this case, multiple legal documents may be progressively queried for the content expressed at a certain shared node, and their meanings, reasons, and the main points of the judgment may be displayed to illustrate the analysis results of the current legal documents when inputting them.

[0064] Among them, overlapping presentation refers to two or more legal documents using the same expression to describe the same legal facts or elements. For example, multiple judgments use the term "marital property" but the specific assets referred to may differ slightly or be completely identical. In this case, it is necessary to check the relevant content of multiple legal documents to see if the differences are conflicting and how the differences are presented.

[0065] S32, based on the probability distribution of the current progressive and overlapping representation methods, form a document path, and based on the data ownership of the document path, obtain the similarity set and conflict set of the current shared node under the data ownership.

[0066] The probability distributions of the aforementioned progressive and overlapping expressions are set based on the frequency of occurrence of the corresponding words in the progressive and overlapping expressions, and are set using the inverse document probability of the corresponding words to illustrate the probability of different word descriptions appearing in the corresponding legal documents. Then, this probability is used to describe the probability of a specific progressive expression and the probability of repeatedly describing a certain legal fact in legal documents related to historical data.

[0067] Meanwhile, the document path represents the path of the evolution and citation of legal facts under progressive and overlapping presentation methods. Then, the data belonging to the parts describing the evolution and citation of the case is identified, and the cosine similarity is used to calculate the words in the corresponding paragraphs of the current shared node. During the calculation, the words of the shared node need to be represented as word vectors, and the current shared node is clustered with other shared nodes. When the cosine similarity is closer to 1, it indicates that there is a high degree of correlation between the current paragraph and the shared node. The parts that have obvious conflicts with the information expressed by the current shared node are identified to identify the conflict set.

[0068] The data attribution of the aforementioned document path is achieved by comparing the descriptions of multiple shared nodes for a certain legal fact under the overlapping representation method, using element types or legal document dictionaries to check whether the corresponding descriptions are conflicting descriptions. If they are conflicting descriptions, the shared nodes are marked. As for the progressive representation method, it compares the multiple texts in the shared nodes about the progression of the legal fact with the texts of other shared nodes to see if there is a conflict at a certain position in the progression, and marks it accordingly. The data marked with conflict is then placed into the conflict set.

[0069] As for the similarity set, the cosine similarity is calculated in the form of word vectors based on the advancement process and corresponding description between multiple shared nodes. The part with a cosine similarity value greater than 0.6 is selected and the corresponding data is used as the similarity set to identify the identification comparison between multiple sets of shared nodes after multiple legal documents are input.

[0070] The parameter interaction mode described above is used to illustrate the process of clustering the current shared node with other shared nodes in a progressive and overlapping manner, thereby describing the interaction process of multiple shared nodes.

[0071] S33, perform cluster analysis on the similarity set and conflict set corresponding to the current shared node, map the cluster center of each cluster to each shared node after analysis, and use the case type described by each shared node after mapping as the analysis result of the current legal document.

[0072] Preferably, the implementation of step S32 further includes: S321, extracting path nodes based on the probability values ​​of the current progressive and overlapping representation methods. The path node selects the main turning point in the vocabulary of the progressive representation method by its probability value. For example, if its probability value is greater than 0.3 or greater than the average probability of multiple words in the progressive representation method in historical data for vocabulary progressive description, it is used as the basis for its segmentation turning point. This division method can find words that frequently appear in the inverse document probability as multiple advancement processes on the vocabulary segmentation path to illustrate the path branches with similarity under the overall advancement, so as to complete the rapid comparison and sorting of legal documents.

[0073] As for the path nodes of the overlapping expression mode, the words that describe a certain legal fact multiple times in a row are used as its path nodes. For example, words that describe a single legal fact three or more times are used as its path nodes to illustrate the words that describe a single legal fact multiple times. Based on the obtained path nodes, the multiple texts corresponding to the progressive expression mode and the overlapping expression mode are further segmented.

[0074] S322, classify the data ownership of each path node, and connect each path node according to the feature type under the data ownership.

[0075] S323 performs similarity comparison and conflict detection on each path node to generate a similarity set and a conflict set under the corresponding data ownership.

[0076] Here, similarity comparison involves converting path nodes into word vectors, calculating cosine similarity, and then selecting the corresponding path nodes to add to the similarity set.

[0077] Conflict detection uses a legal dictionary to check whether the descriptions in the path nodes conflict. If there is a conflict, it is added to the conflict set. Also, if there are differences in the path nodes during the progression, they are added to the corresponding conflict set. This allows for centralized processing of the different descriptions in multiple sets of legal documents.

[0078] Preferably, in step S33, when performing cluster analysis, density clustering is used for similar sets. Density clustering emphasizes the combination of similar patterns, which can identify similar patterns that are described step by step for a legal fact. In density clustering, the cosine similarity of the similar set when the corresponding shared node data is input is used as the neighborhood radius. For example, the part with a cosine similarity greater than 0.85 is used as its neighborhood radius to obtain multiple data combinations with strong correlation. After traversing all nodes, they are placed into the corresponding neighborhood radius, and the number of nodes clustered within the neighborhood radius is set to 5 to obtain multiple clusters for density clustering. Then, the cluster centers are mapped to the shared nodes, and the shared nodes are labeled with the corresponding case type to complete the analysis of the current legal document.

[0079] As for the conflict set, hierarchical clustering is used to display the path of different processing forms under the same legal facts. The relevant data in the conflict set can also be calculated by the word vectors of all nodes. This distance can be obtained by Euclidean distance. The two data with the smallest distance between the word vectors are combined in turn and iterated until the average value of the cluster exceeds the preset threshold to obtain multiple clusters. After mapping these data, the description of the case type corresponding to the conflict set is completed to illustrate the multiple cluster images presented by the current legal documents under the corresponding case type.

[0080] Preferably, the preset threshold used for conflict sets will be calculated based on the Euclidean distance of its clusters, and the preset threshold will be set with the value of the normalized Euclidean distance, which is the average value of historical data in the clusters.

[0081] The final output analysis results will include various descriptive information such as conflict and similar shared node indexes, case types, causes, and element types. This will facilitate the rapid processing of document content based on the descriptions of similar and conflicting parts when comparing multiple legal documents.

[0082] Preferably, the aforementioned case types are used to illustrate the multiple clustering forms that different case types can correspond to after clustering similar sets and conflict sets, quantify the mapping and association description between different case types and data ownership, shared nodes, and element types, and assist in the judicial management of their legal facts.

[0083] Meanwhile, for the currently input legal documents, if there are scenarios where the main structures are similar, clustering anomaly judgment is performed based on the current input structural scenario to obtain whether there are differences in the input of similar documents, and the status of each difference point is marked to the shared node corresponding to each legal document.

[0084] For example, labels can be set in a differential form based on the content of the judgment document, such as the case description and the culpability. The differential labels can be used to explain the word segmentation sharing status between each shared node. The distribution of word segmentation sharing status among the shared nodes can be used to update the status of the legal document input in the current solution.

[0085] In one embodiment of the present invention, during the processing of step S4, it is necessary to process some of the clustered data in step S3, such as extracting the differences between the input legal documents in the similarity set and the conflict set among the multiple currently input legal documents, and identifying the form of these difference nodes.

[0086] like Figure 7As shown, the implementation of step S4 includes: S41, based on the differences of any data in the same cluster in the analysis results of the current legal documents, extract the difference nodes, associate and map the difference nodes with the shared nodes, and determine the matching degree between the shared nodes and the difference nodes.

[0087] S42. According to the access order of the shared nodes, compare the matching degree of any two different nodes. If the matching degree of any two different nodes is greater than the preset matching degree threshold, then mark the corresponding different node as a candidate node that can be merged.

[0088] S43, identify whether the candidate nodes to be merged are parallel nodes. If they are not parallel nodes, merge the candidate nodes to be merged into the difference nodes between the shared nodes. If they are parallel nodes, output the difference nodes between the shared nodes in sequence according to the matching degree value.

[0089] Preferably, in step S41, the two data sets in the similar set and conflict set in any cluster are compared. The differences that exist during the comparison are used as difference nodes. After mapping the relevant content of the difference nodes to the shared nodes, a multi-data association structure is formed.

[0090] Preferably, the matching degree mentioned above represents the similarity between the words in the difference node and the words expressed by the shared node. In this case, the word vector form is used. After the extracted difference node is converted into a word vector, the similarity between the current difference node and the content described in the original legal document is calculated. This similarity will use cosine similarity.

[0091] Preferably, the current access order of the shared nodes represents the order in which the shared nodes are normally viewed. The resulting differences are identified, such as the order of the factual elements chain and the adjudication rule chain corresponding to the shared nodes. This order is used as the current access order of the shared nodes. Each difference node is viewed in turn, and the matching degree of the difference nodes after mapping is determined. After determining whether the identified pairs of difference nodes are in a parallel or merged state, the difference nodes are merged and filtered to obtain the output difference nodes.

[0092] The above method for identifying parallel nodes is based on semantic recognition or extraction using relevant dictionaries in legal documents. The preset matching degree is set based on the average matching degree of historical data when merging related texts in legal documents. If the current matching degree value is not greater than the preset matching degree threshold, it means that the two differing texts are not highly related to the content described by the original shared node. These two differing nodes need to be retained as content to be viewed later when reviewing legal documents. If both are greater than the preset matching degree threshold, it means that they can be merged, and their differences can be processed as information of the same category.

[0093] In one embodiment of the present invention, when conducting transactions with each shared node, the current aggregated shared nodes are obtained mainly in the form of smart contracts, such as verifying the applicant's identity, asymmetric encryption of identity, submitting a transaction application, canceling transaction content, and transmitting transaction data. The aggregated shared nodes, based on their differences and the connection relationships between shared nodes, output relevant legal documents in the form of paths to complete the configuration and processing of legal documents in various scenarios.

[0094] The final output shared nodes can represent shared recognition queries for single texts and combined texts, to filter legal texts relevant to the current needs from legal documents under various conditions and scenarios, complete the legal text shared recognition processing, and promote the efficiency and reliability of legal text utilization.

[0095] In step S5, the relevant data of the difference node is synchronized to the blockchain storing the shared node to ensure that the data of the difference node and the shared node comply with the business rules. Then, the relevant content of the difference node is used to update the relative state of the shared node to pass the influence of the difference node across the shared node, avoid isolated processing, and improve the flexibility of the legal document processing process according to the connection method between the difference node and the adjacent shared node.

[0096] Therefore, the implementation of step S5 also includes connecting the difference node with the adjacent shared node in the form of key-value pairs, and recording the influence path of the difference node on the shared node.

[0097] Define the constraints on the differential nodes and adjacent shared nodes under the influence path, and based on the constraints of the differential nodes, identify the update status of the adjacent shared nodes when the description information of the differential nodes changes.

[0098] The updated neighboring shared nodes are input into the blockchain to complete the node transaction process for the shared nodes.

[0099] The key-value pair format described above uses the difference node as the index of the key-value pair, stores adjacent shared nodes as key-value pairs, and then records the multiple shared nodes connected to the difference node, treating these shared nodes as the influencing path.

[0100] Preferably, when processing the difference nodes in the impact path, it is necessary to check the constraints of adjacent shared nodes and the difference nodes to determine the number of adjacent shared nodes that need to be processed. For example, if there are difference nodes under custody, then the adjacent shared nodes need to include content that can be directly described and related to the difference nodes, such as marital property. Otherwise, it is considered that the current difference node and the adjacent shared nodes cannot be connected, and their connection will affect the consistency of the data described by the adjacent shared nodes.

[0101] In other words, when the extracted difference node is connected to the adjacent shared node, its description must correspond to the logical relationship under legal document processing. Only when there is a logical relationship between the adjacent shared node and the difference node, it is necessary to update the data status of the current shared node according to the change of the difference node data, so as to adjust the configuration of the relevant data of the shared node.

[0102] If there is no logical relationship with adjacent shared nodes, that is, if there is no connection with adjacent shared nodes, the state of the current different node is projected onto the corresponding shared node to adjust the data update status of the current shared node.

[0103] The constraints between different nodes and adjacent shared nodes can be obtained by viewing the corresponding logical relationships, and the corresponding logical relationships can be output as constraints.

[0104] If the data of the difference node changes, that is, when multiple legal documents are entered and the data extracted from the current adjacent shared node changes, check whether the corresponding data of the adjacent shared node has changed. If the logical relationship between the difference node and the adjacent shared node is still satisfied at this time, then update the relevant data of the difference node to the data stored in the adjacent shared node to complete the processing of the relevant data of the shared node.

[0105] It should be noted that the logical relationship between the difference nodes will be explained by the dependency relationship between the corresponding data in the legal knowledge base and the relative conflict relationship, so as to indicate whether there is a relationship between the difference nodes and the adjacent shared nodes.

[0106] Preferably, when connecting difference nodes with adjacent shared nodes using key-value pairs, the implementation further includes: calculating the Euclidean distance between the difference node and its adjacent shared nodes in the time dimension based on the time tag of each difference node, comparing the Euclidean distances of all difference nodes with their adjacent shared nodes in the time dimension, and determining the connection order of the difference nodes and their adjacent shared nodes. After obtaining the Euclidean distances in the time dimension, they are sorted sequentially from smallest to largest based on the distance value to identify the update speed of the difference nodes and their adjacent shared nodes; and the updated content is used to assist subsequent staff in quickly finding relevant content and corresponding differences in the current text when searching for corresponding legal documents, thereby improving the readability and efficiency of legal document interpretation.

[0107] At this point, based on the extraction time of each difference node, the time dimension of the difference node at the time of extraction is compared with the extraction time of adjacent shared nodes to determine whether the relevant stored data of adjacent shared nodes has been updated during multiple data inputs. This facilitates subsequent verification of the update status of adjacent shared nodes. The larger the Euclidean distance between the time point of the difference node and the time point of the adjacent shared node, the larger the update interval of the data stored in the corresponding shared node, indicating untimely updates and potential data consistency issues. If the value is small, it indicates that the adjacent shared node updates frequently, suggesting an active update state and more reliable data. At this point, the data stored in the shared node is further identified based on its relative time point to illustrate the update status of the shared node after the current input of legal documents. This provides a data foundation for subsequent node transaction processes and data update management, ultimately achieving timely updates and iterative aggregation of legal document analysis and management.

[0108] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A method for sharing legal document data based on data analysis, characterized in that, include: S1, cut legal document text according to part-of-speech boundaries, divide legal documents into multiple shared nodes according to the part-of-speech of legal elements, and record the embedding dimension of each shared node; S2, perform structural analysis on each shared node, trace the data source of the shared node based on the integration relationship of each piece of information within the shared node, and extract the structural data corresponding to each shared node; S3 performs interactive analysis on the structural data corresponding to each shared node, obtains the parameter interaction pattern between each shared node, and outputs the analysis results of the current legal document in the form of clustering similar sets and conflict sets after multiple text inputs based on the parameter interaction pattern. S4. Based on the analysis results of the current legal documents, perform anomaly clustering detection on the input legal documents, and identify the difference nodes between the shared nodes according to the access order of each shared node. S5 connects the identified difference nodes with adjacent shared nodes, sets up node transaction processes for shared nodes, and completes the configuration of each shared node; Step S3 can be implemented in the following ways: S31, View the legal facts of shared nodes in the structural data, perform interactive analysis on each shared node based on the legal facts, and obtain the progressive and overlapping expression methods in scenarios with similar word meanings; The progressive expression method refers to the gradual unfolding and deepening of a legal fact or element in different documents; Overlapping representations refer to two or more legal documents using the same expression to describe the same legal fact or element; S32, Based on the probability distribution of the current progressive and overlapping representation methods, form a document path, and based on the data ownership of the document path, obtain the similarity set and conflict set of the current shared node under the data ownership. Document path refers to the evolution and citation of legal facts in a case under progressive and overlapping presentation methods; S33, perform cluster analysis on the similarity set and conflict set corresponding to the current shared node, map the cluster center of each cluster to each shared node after analysis, and use the case type described by each shared node after mapping as the analysis result of the current legal document; The implementation of step S32 also includes: S321, Extract path nodes based on the probability values ​​of the current progressive and overlapping representation methods; S322, classify the data ownership of each path node, and connect each path node according to the feature type under the data ownership; S323 performs similarity comparison and conflict detection on each path node to generate a similarity set and a conflict set under the corresponding data ownership.

2. The legal document data sharing method based on data analysis according to claim 1, characterized in that, The implementation methods for step S1 include: S11, When the requester accesses the legal document, the semantic analysis of the current legal document is performed, and the part of speech of the legal elements of the current legal document is set with the preset part of speech boundary as the word segmentation condition; S12, using the data corresponding to the part-of-speech of the legal elements in the current legal document as the initial node, defining the subordinate components of each initial node under the part-of-speech boundary, identifying the subordinate sub-words of each initial node, and performing position embedding, sub-word embedding and sentence embedding in sequence based on the sentence and position of the subordinate sub-words; S13, use the labels of each initial node to perform label embedding, regard the initial nodes after embedding as the output shared nodes, and use the embedded data as the embedding dimension of each shared node.

3. The legal document data sharing method based on data analysis according to claim 1, characterized in that, The implementation of step S2 also includes: S21. Based on the legal structure, factual structure, evidentiary structure, and temporal structure corresponding to the current shared node, determine whether the current shared node belongs to the vertical integration, horizontal integration, or dynamic integration mode. S22, Based on the current integration method of shared nodes, generate structural fingerprints, perform path backtracking based on the structural fingerprints of each shared node, and identify the source analysis network corresponding to each shared node; S23, explore all associated paths of the current shared node in the source analysis network, and output the associated paths of the current shared node as the structural data of each shared node.

4. The legal document data sharing method based on data analysis according to claim 3, characterized in that, The implementation method of step S21 is represented as follows: S211, If ​​the current shared node belongs to vertical integration, then based on the multiple sets of data of the shared node in vertical integration, the shared nodes corresponding to the vertical integration are connected in a hierarchical manner to form a vertical hierarchical structure. S212, If the current shared node belongs to horizontal integration, then a horizontal integration structure is formed based on the logical complementarity between multiple legal texts of the application; S213, If the current shared nodes are in dynamic integration, a dynamic integration structure is formed based on the time sequence advancement between shared nodes under the same integration method.

5. The legal document data sharing method based on data analysis according to claim 3, characterized in that, The implementation methods of step S22 include: Obtain the connection methods of shared nodes under different integration relationships, and identify the feature type based on multiple sets of shared nodes corresponding to the connection methods; Obtain the feature types of shared nodes under the corresponding group, and set the source tracing paths between shared nodes based on the pointing relationship of feature types in shared nodes, and combine the source tracing paths to form the source tracing analysis network; If there are multiple intersecting parts between the tracing paths, the shared nodes of the intersecting parts are integrated based on their pointing relationships and referencing methods, and the integrated shared nodes point to other shared nodes.

6. The legal document data sharing method based on data analysis according to claim 1, characterized in that, Step S4 can be implemented in the following ways: S41. Based on the differences in any data within the same cluster in the analysis results of the current legal documents, extract the difference nodes, associate and map the difference nodes with the shared nodes, and determine the matching degree between the shared nodes and the difference nodes. S42, according to the access order of the shared nodes, compare the matching degree of any two different nodes. If the matching degree of any two different nodes is greater than the preset matching degree threshold, then mark the corresponding different node as a candidate node that can be merged. S43, identify whether the candidate nodes to be merged are parallel nodes. If they are not parallel nodes, merge the candidate nodes to be merged into the difference nodes between the shared nodes. If they are parallel nodes, output the difference nodes between the shared nodes in sequence according to the matching degree value.

7. The method for sharing legal document data based on data analysis according to claim 1, characterized in that, The implementation of step S5 also includes: Connect the differing node to the adjacent shared node using key-value pairs to record the influence path of the differing node on the shared node; define the constraints of the differing node and the adjacent shared node under the influence path; based on the constraints of the differing node, identify the update status of the adjacent shared node when the description information of the differing node changes; input the updated adjacent shared node into the blockchain to complete the node transaction process of the shared node.

8. A method for sharing legal document data based on data analysis according to claim 7, characterized in that, When connecting differing nodes with adjacent shared nodes using key-value pairs, the implementation methods also include: Based on the time label of each difference node, calculate the Euclidean distance between the difference node and its adjacent shared nodes in the time dimension, and compare the Euclidean distances between all difference nodes and their adjacent shared nodes in the time dimension to determine the connection order between the difference nodes and their adjacent shared nodes.

Images