A case template structure creation system and method based on large language model

Abstract

The present invention provides a system and method for creating a case template structure based on a large language model: a behavior fact extraction creation module is used to create and extract the factual content of the behavior facts constituting a certain case based on the case constituent elements that constitute the case; a plot identification creation module is used to create identification standards for the plot identification that constitute the case and qualitative conclusions of the plot identification corresponding to the identification standards, and to create an association between the factual content of the case and the identification standards and the qualitative conclusions of the plot identification; a case element creation module is used to create an association between the case constituent elements and case component elements that constitute the case and the case conclusion, and to create an association support relationship between the qualitative conclusions of the plot identification of the case and the case constituent elements; a case graph template creation module is used to create a case graph template of the case based on the case constituent elements of the case and using a knowledge graph; a large model construction module is used to train and learn the large language model using industry cases to obtain an industry-specific large model.

Description

Technical Field

[0001] The present invention relates to the technical field of large language models and knowledge graphs, and in particular to a system and method for creating a case template structure based on a large language model. Background Art

[0002] The patent previously applied by the applicant mentioned a case template constructed based on the knowledge graph principle. This case template is relatively simple and can only extract behavioral facts from the case file materials to form a case graph for reference by case handlers, so that they can understand the general situation of the case as soon as possible. However, the existing case template structure does not have a setting for plot identification. Case handlers need to make plot identification based on the case graph, and cannot automatically realize plot identification based on behavioral facts. Based on this, the present invention designs a new case template structure that can set plot identification, and then quickly identify the plot based on the behavioral facts to obtain a qualitative conclusion of the plot identification, which is more conducive to case handlers to understand the situation of the case in a timely manner and improve the work efficiency of case handlers. Summary of the invention

[0003] In view of the problems and shortcomings of the prior art, the present invention provides a system and method for creating a case template structure based on a large language model.

[0004] The present invention solves the above technical problems through the following technical solutions:

[0005] The present invention provides a case template structure creation system based on a large language model, which is characterized in that it includes a behavior fact extraction creation module, a plot identification creation module, a case element creation module, a case graph template creation module and a large model construction module;

[0006] The behavior fact extraction and creation module is used to create and extract the factual content of the behavior facts constituting a certain cause of action based on the case constituent elements constituting the cause of action;

[0007] The scenario identification creation module is used to create the identification criteria for the scenario identification of the case and the qualitative conclusions of the scenario identification corresponding to the identification criteria, and to create the association between the factual content of the case and the identification criteria and the qualitative conclusions of the scenario identification. Different factual contents of the case are associated with different identification criteria and qualitative conclusions of the scenario identification.

[0008] The case element creation module is used to create the association relationship between the case elements and case component elements that constitute the case and the case conclusion, and to create the association support relationship between the qualitative conclusion of the circumstances of the case and the case elements. Different qualitative conclusions of the circumstances of the case are associated with different case elements, and the different case elements are associated with different case conclusions.

[0009] The case graph template creation module is used to create a case graph template of the case based on the case constituent elements of the case and using the knowledge graph principle;

[0010] The large model construction module is used to use the accumulated industry cases to train and learn the large language model to obtain an industry-specific large model. The industry cases all have associated extracted behavioral facts, qualitative conclusions of circumstances after circumstances identification, and case constituent elements. The industry-specific large model is used to extract behavioral facts for individual cases. The extracted behavioral facts are used to fill in the corresponding case cause graph template, and the extracted behavioral facts are matched one by one with the associated circumstances identification standards to match whether a certain identification standard is met. When a certain identification standard is met, a corresponding qualitative conclusion of circumstances identification is given, and the given qualitative conclusion of circumstances identification is associated with the case constituent elements to support the corresponding case cause conclusion.

[0011] The present invention also provides a method for creating a case template structure based on a large language model, which is characterized in that it includes the following steps:

[0012] Relying on the case constituent elements that constitute a certain cause of action, create and extract the factual content of the behavioral facts that constitute the cause of action;

[0013] Establish the identification criteria for the circumstances of the case and the qualitative conclusions of the circumstances corresponding to the identification criteria, and establish the relationship between the factual content of the case and the identification criteria and the qualitative conclusions of the circumstances, so that different factual contents of the case are associated with different identification criteria and qualitative conclusions of the circumstances;

[0014] Establish the relationship between the case elements and case component elements that constitute the case and the case conclusion, and establish the relationship between the qualitative conclusion of the circumstances of the case and the case elements. Different qualitative conclusions of the circumstances of the case are associated with and support different case elements, and different case elements are associated with different case conclusions.

[0015] Based on the case constituent elements of the case and using the knowledge graph principle, a case graph template of the case is created;

[0016] The large language model is trained and learned using the accumulated industry cases to obtain an industry-specific large model. All industry cases have associated extracted behavioral facts, qualitative conclusions on circumstances after circumstances identification, and case constituent elements. The industry-specific large model is used to extract behavioral facts from the behavioral facts of individual cases. The extracted behavioral facts are used to fill in the corresponding case cause graph template, and the extracted behavioral facts are matched one by one with the associated circumstances identification standards to match whether a certain identification standard is met. When a certain identification standard is met, a corresponding qualitative conclusion on circumstances identification is given, and the given qualitative conclusion on circumstances identification is associated with the case constituent elements to support a corresponding case cause conclusion.

[0017] The positive and progressive effect of the present invention is that the present invention designs a new case template structure that can set the plot identification, and then can quickly identify the plot based on the behavioral facts to obtain a qualitative conclusion of the plot identification, which is more conducive to the case handlers to understand the case situation in a timely manner and improve the work efficiency of the case handlers. The case template structure designed by the present invention is a closed-loop case template structure consisting of case elements, behavioral facts and plot identification. The case elements determine the behavioral facts, the behavioral facts determine the plot identification, and the plot identification supports the case elements. On the basis of the existing case template, plot identification and case elements are added. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A structural schematic diagram of a case template structure creation system according to a preferred embodiment of the present invention.

[0019] Figure 2 A schematic diagram of a case template structure according to a preferred embodiment of the present invention.

[0020] Figure 3 A schematic diagram of a case graph template created according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0022] like Figure 1-3As shown, this embodiment provides a case cause template structure creation system based on a large language model, which includes a behavior fact extraction creation module 1, a scenario identification creation module 2, a case element creation module 3, a case cause graph template creation module 4, a large model construction module 5, a case cause graph template testing module 6 and a case cause graph template management module 7.

[0023] The behavior fact extraction creation module 1 is used to create and extract the factual content of the behavior facts constituting a certain cause of action based on the case constituent elements constituting the crime of smuggling ordinary goods. For example, the behavior fact extraction creation module 1 can create and extract the factual content of the behavior facts constituting the crime of smuggling ordinary goods based on the case constituent elements constituting the crime of smuggling ordinary goods. When a case file material corresponds to the crime of smuggling ordinary goods, the created factual content of the behavior facts of the crime of smuggling ordinary goods is used to specifically extract the specific behavior facts in the case file material.

[0024] The scenario identification creation module 2 is used to create the identification standards for the scenario identification of the case and the qualitative conclusions of the scenario identification corresponding to the identification standards, as well as to create the relationship between the factual content of the case and the identification standards and the qualitative conclusions of the scenario identification. Different factual contents of the case are associated with different identification standards and qualitative conclusions of the scenario identification.

[0025] For example, the scenario identification creation module 2 can create the identification criteria for the scenario identification of the crime of smuggling ordinary goods and the qualitative conclusions of the scenario identification corresponding to the identification criteria, as well as the relationship between the factual content of the crime of smuggling ordinary goods and the identification criteria and the qualitative conclusions of the scenario identification. Scenario identification is to characterize the extracted behavior facts and give a qualitative conclusion of the scenario identification.

[0026] For example, if there is only the circumstance of smuggling ordinary goods but no circumstance of surrendering oneself, or if there is the circumstance of smuggling ordinary goods and circumstance of surrendering oneself, different circumstance identification standards will correspond, and different qualitative conclusions on circumstance identification will be given in the end.

[0027] The case element creation module 3 is used to create the relationship between the case elements and case component elements that constitute the case and the case conclusion, as well as to create the relationship between the qualitative conclusion of the circumstances of the case and the case elements. Different qualitative conclusions of the circumstances of the case are associated with and support different case elements, and the different case elements supported are associated with different case conclusions.

[0028] According to the functions of the behavior fact extraction creation module 1, the situation identification creation module 2 and the case element creation module 3, the case element, behavior facts and situation identification constitute a closed-loop case template structure (see Figure 2), the case elements determine the facts of the behavior, the facts of the behavior determine the determination of the circumstances, and the determination of the circumstances supports the case elements. The case template structure stipulates which facts of a crime are extracted, which facts of the behavior are associated with which circumstances determination standards, and which circumstances determination standards are associated with which case elements.

[0029] The case graph template creation module 4 is used to create a case graph template for the case based on the case constituent elements of the case and using the knowledge graph principle. The case graph template determines which behavioral facts are extracted, and these behavioral facts are related to the determination of the circumstances.

[0030] For example, based on the case elements of the crime of smuggling ordinary goods, and using the knowledge graph principle, a graph template for the crime of smuggling ordinary goods is created (see Figure 3 ), after receiving the relevant case files, the industry-specific large model can be used to extract the behavioral facts in the case files, and the extracted behavioral facts can be filled into the map template of the crime of smuggling ordinary goods to form a case map of the crime of smuggling ordinary goods, so that case handlers can easily view the content of each behavioral fact and understand the case situation in a timely manner.

[0031] The big model construction module 5 is used to use the accumulated industry cases to train and learn the big language model to obtain an industry-specific big model. The industry cases all have associated extracted behavioral facts, qualitative conclusions of circumstances after circumstances identification, and case constituent elements; the trained and learned industry-specific big model is used to extract behavioral facts from the behavioral facts of individual cases, and the extracted behavioral facts are used to fill in the corresponding case cause map template to form an individual case map, and the extracted behavioral facts are matched one by one with the related circumstances identification identification standards to match whether a certain identification standard is met, and when a certain identification standard is met, a corresponding qualitative conclusion of circumstances identification is given, and the given qualitative conclusion of circumstances identification is associated with the case constituent elements to support the corresponding case cause conclusion.

[0032] The case cause graph template testing module 6 is used to test the created case cause graph template of the case cause using the case of the case cause, use the industry-specific large model to extract the behavioral facts of the case, and fill the extracted behavioral facts into the case cause graph template of the case cause and map the extracted behavioral facts with the original content source in the case, and test and evaluate the advantages and disadvantages of the created case cause graph template of the case cause based on the extraction and filling results and the mapping relationship.

[0033] The case graph template testing module 6 is also used to use the case of the case to fully test the first version of the case graph template created for the case, and supports testing the second version and subsequent versions of the case graph template created for the case with any node as the starting point.

[0034] The case cause graph template testing module 6 is also used to support separate incremental testing of only the newly added branch node or the modified branch node when a branch node is added or a branch node is modified in the case cause graph template of the case.

[0035] The case cause graph template management module 7 is used to perform version management on the case cause graph template, and when a branch node is added or a branch node is modified in the case cause graph template of the case, the newly added case cause graph template or the modified case cause graph template is saved as the latest version of the case cause graph template, and when the case cause graph template of the case is used, the latest version of the case cause graph template is used as the target case cause graph template.

[0036] This embodiment also provides a method for creating a case template structure based on a large language model, which includes the following steps:

[0037] Create factual content that extracts the behavioral facts that constitute a cause of action.

[0038] Create the identification standards for the circumstances that constitute the case and the qualitative conclusions on the circumstances that correspond to the identification standards, as well as create the relationship between the factual content of the case and the identification standards and the qualitative conclusions on the circumstances. Different factual contents of the case are associated with different identification standards and qualitative conclusions on the circumstances.

[0039] Create the relationship between the case constituent elements and case component elements that constitute the case and the case conclusion, and create the relationship between the qualitative conclusion of the circumstances determination of the case and the case constituent elements. Different qualitative conclusions of the circumstances determination of the case are associated with and support different case constituent elements, and the different case constituent elements supported are associated with different case conclusions.

[0040] Based on the case constituent elements of the case and using the knowledge graph principle, a case graph template for the case is created.

[0041] The large language model is trained and learned using the accumulated industry cases to obtain an industry-specific large model. All industry cases have associated extracted behavioral facts, qualitative conclusions on circumstances after circumstances identification, and case constituent elements. The industry-specific large model is used to extract behavioral facts from the behavioral facts of individual cases. The extracted behavioral facts are used to fill in the corresponding case cause graph template, and the extracted behavioral facts are matched one by one with the associated circumstances identification standards to match whether a certain identification standard is met. When a certain identification standard is met, a corresponding qualitative conclusion on circumstances identification is given, and the given qualitative conclusion on circumstances identification is associated with the case constituent elements to support a corresponding case cause conclusion.

[0042] The case of this case is used to test the created case cause graph template of this case. The behavioral facts of the case are extracted using an industry-specific large model. The extracted behavioral facts are filled into the case cause graph template of this case and the extracted behavioral facts are mapped with the original content source in the case. The advantages and disadvantages of the created case cause graph template of this case are tested and evaluated based on the extraction and filling results and the mapping relationship.

[0043] The first version of the case graph template created for the case is fully tested using the case of the case, and the second version and subsequent versions of the case graph template created for the case can be tested starting from any node.

[0044] If a branch node is added or a branch node is modified in the case graph template of the case, it supports separate incremental testing of only the newly added branch node or the modified branch node.

[0045] The case cause graph template is version managed, and when a branch node is added or a branch node is modified in the case cause graph template of the case, the newly added case cause graph template or the modified case cause graph template is saved as the latest version of the case cause graph template, and when the case cause graph template of the case is used, the latest version of the case cause graph template is used as the target case cause graph template.

[0046] Although the specific embodiments of the present invention are described above, those skilled in the art should understand that these are only examples, and the protection scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but these changes and modifications all fall within the protection scope of the present invention.

Claims

1. A case template structure creation system based on a large language model, characterized in that: It includes a behavior fact extraction creation module, a scenario identification creation module, a case element creation module, a case cause graph template creation module and a large model construction module; The behavior fact extraction and creation module is used to create and extract the factual content of the behavior facts constituting a certain cause of action based on the case constituent elements constituting the cause of action; The scenario identification creation module is used to create the identification criteria for the scenario identification of the case and the qualitative conclusions of the scenario identification corresponding to the identification criteria, and to create the association between the factual content of the case and the identification criteria and the qualitative conclusions of the scenario identification. Different factual contents of the case are associated with different identification criteria and qualitative conclusions of the scenario identification. The case element creation module is used to create the case elements that constitute the case cause and the relationship between the case elements and the case cause conclusion, and to create the relationship between the qualitative conclusion of the circumstances of the case and the elements of the case. Different qualitative conclusions of the circumstances of the case are associated with different case elements, and the different case elements supported are associated with different case cause conclusions. The case graph template creation module is used to create a case graph template of the case based on the case constituent elements of the case and using the knowledge graph principle; The large model construction module is used to use the accumulated industry cases to train and learn the large language model to obtain an industry-specific large model. The industry cases all have associated extracted behavioral facts, qualitative conclusions of circumstances after circumstances identification, and case constituent elements. The industry-specific large model is used to extract behavioral facts for individual cases. The extracted behavioral facts are used to fill in the corresponding case cause graph template, and the extracted behavioral facts are matched one by one with the associated circumstances identification standards to match whether a certain identification standard is met. When a certain identification standard is met, a corresponding qualitative conclusion of circumstances identification is given, and the given qualitative conclusion of circumstances identification is associated with the case constituent elements to support the corresponding case cause conclusion.

2. The case template structure creation system based on a large language model as claimed in claim 1, characterized in that: The system also includes a case graph template testing module; The case cause graph template testing module is used to test the created case cause graph template of the case cause using the case of the case cause, extract the behavioral facts of the case using an industry-specific large model, fill the extracted behavioral facts into the case cause graph template of the case cause, and map the extracted behavioral facts with the original content source in the case, and test and evaluate the advantages and disadvantages of the created case cause graph template of the case cause based on the extraction and filling results and the mapping relationship.

3. The case template structure creation system based on a large language model as claimed in claim 2, characterized in that: The case cause graph template testing module is also used to use the case of the case cause to fully test the first version of the case cause graph template created for the case cause, and supports testing the second version and subsequent versions of the case cause graph template created for the case cause from any node as the starting point.

4. The case template structure creation system based on a large language model as claimed in claim 2, characterized in that: The case cause graph template testing module is also used to add a branch node or modify a branch node in the case cause graph template of the case cause, and supports separate incremental testing of only the newly added branch node or the modified branch node.

5. The case template structure creation system based on a large language model as claimed in claim 2, characterized in that: The system also includes a case graph template management module; The case cause graph template management module is used to perform version management on the case cause graph template, and when a branch node is added or a branch node is modified in the case cause graph template of the case, the newly added case cause graph template or the modified case cause graph template is saved as the latest version of the case cause graph template, and when the case cause graph template of the case is used, the latest version of the case cause graph template is used as the target case cause graph template.

6. A method for creating a case template structure based on a large language model, characterized in that: It includes the following steps: Relying on the case constituent elements that constitute a certain cause of action, create and extract the factual content of the behavioral facts that constitute the cause of action; Establish the identification criteria for the circumstances of the case and the qualitative conclusions of the circumstances corresponding to the identification criteria, and establish the relationship between the factual content of the case and the identification criteria and the qualitative conclusions of the circumstances, so that different factual contents of the case are associated with different identification criteria and qualitative conclusions of the circumstances; Establish the case elements that constitute the case cause and the relationship between the case elements and the case conclusion, and establish the relationship between the qualitative conclusion of the circumstances of the case and the elements of the case. Different qualitative conclusions of the circumstances of the case are associated with and support different case elements, and different case elements are associated with different case conclusions. Based on the case constituent elements of the case and using the knowledge graph principle, a case graph template of the case is created; The large language model is trained and learned using the accumulated industry cases to obtain an industry-specific large model. All industry cases have associated extracted behavioral facts, qualitative conclusions on circumstances after circumstances identification, and case constituent elements. The industry-specific large model is used to extract behavioral facts from the behavioral facts of individual cases. The extracted behavioral facts are used to fill in the corresponding case cause graph template, and the extracted behavioral facts are matched one by one with the associated circumstances identification standards to match whether a certain identification standard is met. When a certain identification standard is met, a corresponding qualitative conclusion on circumstances identification is given, and the given qualitative conclusion on circumstances identification is associated with the case constituent elements to support a corresponding case cause conclusion.

7. The method for creating a case template structure based on a large language model according to claim 6, characterized in that: The case of this case is used to test the created case cause graph template of this case. The behavioral facts of the case are extracted using an industry-specific large model. The extracted behavioral facts are filled into the case cause graph template of this case and the extracted behavioral facts are mapped with the original content source in the case. The advantages and disadvantages of the created case cause graph template of this case are tested and evaluated based on the extraction and filling results and the mapping relationship.

8. The method for creating a case template structure based on a large language model according to claim 7, characterized in that: The first version of the case graph template created for the case is fully tested using the case of the case, and the second version and subsequent versions of the case graph template created for the case can be tested starting from any node.

9. The method for creating a case template structure based on a large language model according to claim 7, characterized in that: If a branch node is added or a branch node is modified in the case graph template of the case, it supports separate incremental testing of only the newly added branch node or the modified branch node.

10. The method for creating a case template structure based on a large language model according to claim 7, characterized in that: The case cause graph template is version managed, and when a branch node is added or a branch node is modified in the case cause graph template of the case, the newly added case cause graph template or the modified case cause graph template is saved as the latest version of the case cause graph template, and when the case cause graph template of the case is used, the latest version of the case cause graph template is used as the target case cause graph template.

Images