Method and device for generating a transcript based on a large model and computer product

Abstract

The application provides a large model-based record generation method, a large model-based record generation device and a computer product, relates to the large model generation technical field, and the large model-based record generation method comprises the following steps: obtaining current case information; generating a plurality of interrogation question points corresponding to the case information, generating interrogation outline information corresponding to the interrogation question points in a piece-by-piece manner, and displaying the interrogation outline information on a smart terminal; in response to a police officer entering a first instruction on the smart terminal, determining first record information corresponding to the interrogation outline information; determining unanswered questions according to the first record information, generating corresponding follow-up question points, and in response to the police officer entering a second instruction on the smart terminal, updating the first record information into target record information; obtaining a judicial expertise opinion and a plurality of proof evidence materials; according to the case information, the target record information, the judicial expertise opinion and the plurality of proof evidence materials, determining corresponding applicable law articles, and generating legal documents containing the applicable law articles.

Description

Technical Field

[0001] This application relates to the field of large model generation technology, and in particular to a method, device and computer product for generating transcripts based on large models. Background Technology

[0002] In the field of judicial case handling, record-keeping is a core step in documenting the interrogation process, securing evidence, and reflecting the facts of the case. Traditional record-keeping methods heavily rely on the personal experience, legal knowledge, and on-the-spot adaptability of the investigating officers. Officers must manually conceive interrogation outlines and key questions based on limited case information, and independently assess the completeness of answers during the questioning process, generating follow-up questions in real time. This process is not only inefficient and time-consuming, but also carries risks such as omissions in questioning due to differences in individual skill levels, insufficient uncovering of key facts, and incomplete evidence chains, potentially affecting the quality of case handling. Summary of the Invention

[0003] The main objective of this invention is to provide a method for generating transcripts based on a large model, which aims to improve the efficiency, standardization, and accuracy of case handling.

[0004] To achieve the above objectives, the present invention provides a record generation method based on a large model, the record generation method based on a large model comprising:

[0005] Obtain current case information;

[0006] Generate several interrogation question points corresponding to the case information, and generate interrogation outline information corresponding to each interrogation question point, and display the interrogation outline information on the smart terminal;

[0007] In response to the police officer's input of a first instruction on the smart terminal, the first transcript information corresponding to each of the interrogation outline information is determined;

[0008] Based on the first transcript information, unanswered questions are identified, corresponding follow-up questions are generated, and in response to the police officer's input of a second command on the smart terminal, the first transcript information is updated to the target transcript information.

[0009] Obtain the forensic appraisal report and multiple pieces of supporting evidence;

[0010] Based on the case information, target transcript information, forensic expert opinion, and multiple evidentiary materials, the corresponding applicable legal provisions are determined, and a legal document containing the applicable legal provisions is generated.

[0011] Optionally, the step of determining the corresponding applicable legal provisions based on the case information, target transcript information, forensic appraisal report, and multiple evidentiary materials, and generating a legal document containing the applicable legal provisions, includes:

[0012] The legal elements of case information are analyzed using a large model to extract a set of key factual features.

[0013] The key fact feature set is semantically associated and matched with the target transcript information to construct a case fact map.

[0014] Based on the aforementioned case fact map, forensic expert opinion, and / or multiple pieces of aforementioned evidentiary material, the integrity of the evidence chain is verified to determine the integrity of the evidence chain.

[0015] Based on the completeness of the evidence chain, a list of applicable legal provisions is generated by matching a preset legal provisions knowledge graph. The list of applicable legal provisions includes at least one applicable legal provision.

[0016] The legal document is generated according to the recommended list of applicable legal provisions and the target transcript information.

[0017] Optionally, the step of generating a recommended list of applicable legal provisions by matching a preset legal knowledge graph according to the completeness of the evidence chain includes:

[0018] Based on the assessment results of the completeness of the chain of evidence, the applicability matching degree of each legal provision is calculated;

[0019] Based on the matching degree, the legal provisions in the preset legal provisions knowledge graph are initially screened;

[0020] The selected legal provisions are subjected to multi-dimensional feature weighting calculations to generate a priority ranking result for the legal provisions.

[0021] Based on the priority ranking of the legal provisions, the top N legal provisions are extracted to form a recommended list of applicable legal provisions, where N is an integer greater than or equal to 1.

[0022] Optionally, the step of generating several interrogation questioning points corresponding to the case information, generating interrogation outline information corresponding to each interrogation questioning point, and displaying the interrogation outline information on the smart terminal includes:

[0023] Determine the case type and cause of action characteristics corresponding to the case information;

[0024] Based on case type and cause of action features, match the corresponding interrogation template framework from the pre-set interrogation template library;

[0025] By performing deep semantic analysis on the case information using a large model, a set of core interrogation elements corresponding to the interrogation template framework is generated.

[0026] The core interrogation elements are organized hierarchically according to the logical structure of the interrogation template framework to generate structured interrogation outline information.

[0027] Based on the display layout specifications of the aforementioned interrogation template framework, the interrogation outline information is dynamically and visually displayed in different areas on the smart terminal interface using an adaptive layout algorithm.

[0028] Optionally, the display layout specification based on the interrogation template framework uses an adaptive layout algorithm to dynamically visualize the interrogation outline information in different areas on the smart terminal interface, including:

[0029] The display layout specifications of the interrogation template framework are analyzed to obtain the display priority configuration and association configuration of each interrogation element;

[0030] Based on the display priority configuration, the interrogation outline information is divided into a main display area and an auxiliary display area, and the association between the main display area and the auxiliary display area is established;

[0031] Based on the configuration of association relationships and the screen characteristics of smart terminals, the optimal layout parameters and display styles of the main display area and the auxiliary display area are dynamically calculated;

[0032] The display is shown on the smart terminal according to the optimal layout parameters and the display style;

[0033] The display content of the main display area and the auxiliary display area are adjusted in real time according to the progress of the interrogation.

[0034] Optionally, the step of determining the first transcript information corresponding to each of the interrogation outline information in response to the police officer's input of a first instruction on the smart terminal includes:

[0035] Continuously collect the answers that police officers input on their smart terminals, which correspond to the information in each interrogation outline;

[0036] Key information elements are extracted from the answer content using deep learning semantic analysis technology;

[0037] The extracted key information elements are logically verified, and the verified key information elements are used to generate the first transcript information in a preset format.

[0038] Optionally, the step of determining unanswered questions based on the first transcript information, generating corresponding follow-up questions, and updating the first transcript information to the target transcript information in response to the police officer's input of a second instruction on the smart terminal includes:

[0039] The first transcript information is analyzed using an integrity detection algorithm to identify missing information points and contradictions.

[0040] Based on the question-and-answer logical relationship, the key factors that need to be confirmed in the points of contradiction are identified.

[0041] A set of follow-up questions corresponding to the missing information points and the key factors is generated through a multi-turn dialogue model;

[0042] The set of follow-up questions is prioritized according to importance and urgency using a sorting algorithm and then displayed on the smart terminal.

[0043] In response to the police officer's input of a second command on the smart terminal, supplementary information corresponding to the set of follow-up questions is generated;

[0044] The supplementary information is fused with the first transcript information to generate the target transcript information.

[0045] Optionally, the method for generating transcripts based on a large model further includes:

[0046] Obtain legal templates and legal document formats;

[0047] The legal document is examined according to the legal template and the legal document format to determine the level of standardization of the legal document;

[0048] If the standardization level is lower than a preset standardization threshold, non-standard content annotations are generated and corresponding modification suggestions are generated;

[0049] In response to the police officer's modification instructions on the smart terminal, the non-standard content of the legal document is adjusted to generate a legal document that meets the standard.

[0050] Furthermore, to achieve the above objectives, the present invention also provides a generation apparatus, the generation apparatus comprising: a memory, a processor, and a large-model-based record generation program stored in the memory and executable on the processor, the large-model-based record generation program being configured to implement the large-model-based record generation method as described above.

[0051] In addition, to achieve the above objectives, the present invention also provides a computer product including the generation apparatus described above.

[0052] This invention, through obtaining current case information, generates several interrogation questioning points corresponding to the case information, and generates interrogation outlines corresponding to each questioning point. These outlines are then displayed on a smart terminal. In response to a police officer's input of a first command on the smart terminal, the invention determines the first transcript information corresponding to each interrogation outline, identifies unanswered questions based on the first transcript information, generates corresponding follow-up questions, and updates the first transcript information to the target transcript information in response to a second command input by the police officer on the smart terminal. Furthermore, it obtains a forensic expert opinion and multiple pieces of evidentiary evidence. Finally, based on the case information, the target transcript information, the forensic expert opinion, and the multiple pieces of evidentiary evidence, it determines the applicable legal provisions and generates a legal document containing the applicable legal provisions. Thus, by automatically generating interrogation outlines, providing real-time assistance in follow-up questioning, verifying the completeness of evidence, and automatically matching applicable legal provisions, the invention effectively improves case-handling efficiency, standardization, and accuracy. Attached Figure Description

[0053] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a schematic diagram of a method for generating transcripts based on a large model according to an embodiment of the present invention;

[0056] Figure 2 This is a schematic diagram of a method for generating transcripts based on a large model, according to another embodiment of the present invention.

[0057] Figure 3 This is a schematic diagram of a method for generating transcripts based on a large model, according to another embodiment of the present invention.

[0058] Figure 4 This is a schematic diagram of a method for generating transcripts based on a large model, according to another embodiment of the present invention.

[0059] Figure 5 This is a schematic diagram of a method for generating transcripts based on a large model according to another embodiment of the present invention;

[0060] Figure 6 This is a schematic diagram of a method for generating transcripts based on a large model, according to another embodiment of the present invention.

[0061] Figure 7 This is a schematic diagram of a method for generating transcripts based on a large model, according to another embodiment of the present invention.

[0062] Figure 8 This is a schematic diagram of a method for generating transcripts based on a large model, according to another embodiment of the present invention.

[0063] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0064] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Well-known modules, units, and their connections, links, communications, or operations are not shown or described in detail. Furthermore, the described features, architectures, or functions can be combined in any way in one or more embodiments. Those skilled in the art should understand that the various embodiments described below are only for illustrative purposes and not for limiting the scope of protection of the present invention. It is also readily understood that the modules, units, or processing methods in the various embodiments described herein and shown in the accompanying drawings can be combined and designed in various different configurations. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0065] The definitions of various terms or methods used in the following embodiments are, except where logically impossible, generally defined as broad concepts that can be implemented under the premise of the content disclosed in the embodiments. Under this understanding, all specific subordinate limitations of the terms or methods should be considered as part of the invention and should not be narrowly interpreted or biased simply because the specification does not disclose such a specific limitation. Similarly, provided that it is logically feasible, the order of the steps in the method is flexible and varied, and all specific subordinate limitations in the broad concepts of various terms or methods fall within the scope of protection of this invention.

[0066] In the field of judicial case handling, record-keeping is a core step in documenting the interrogation process, securing evidence, and reflecting the facts of the case. Traditional record-keeping methods heavily rely on the personal experience, legal knowledge, and on-the-spot adaptability of the investigating officers. Officers must manually conceive interrogation outlines and key questions based on limited case information, and independently assess the completeness of answers during the questioning process, generating follow-up questions in real time. This process is not only inefficient and time-consuming, but also carries risks such as omissions in questioning due to differences in individual skill levels, insufficient uncovering of key facts, and incomplete evidence chains, potentially affecting the quality of case handling.

[0067] Furthermore, drafting legal documents is equally tedious, requiring police officers to synthesize case information, transcripts, expert opinions, and other materials from multiple sources, manually searching and citing relevant legal provisions, which is prone to errors in standardization. Although some auxiliary tools have emerged in the existing technology, their level of intelligence is limited, relying mostly on pre-set fixed templates and rules, making it difficult to adapt to ever-changing case scenarios, and unable to perform deep semantic understanding, intelligent reasoning, and dynamic decision-making. In particular, they have significant shortcomings in generating targeted follow-up questions, verifying the completeness of the evidence chain, and automatically matching applicable legal provisions.

[0068] The main solution of this application embodiment is as follows: by obtaining current case information, generating several interrogation questioning points corresponding to the case information, generating interrogation outline information corresponding to each interrogation questioning point, and displaying the interrogation outline information on a smart terminal, then in response to the police officer's input of a first instruction on the smart terminal, determining the first transcript information corresponding to each interrogation outline information, and determining unanswered questions based on the first transcript information, generating corresponding follow-up questioning points, and in response to the police officer's input of a second instruction on the smart terminal, updating the first transcript information to the target transcript information, then obtaining a forensic appraisal report and multiple evidentiary materials, and finally, based on the case information, target transcript information, forensic appraisal report, and multiple evidentiary materials, determining the corresponding applicable legal provisions, and generating a legal document containing the applicable legal provisions.

[0069] In this embodiment, for ease of description, the generating device will be used as the execution subject in the following description.

[0070] This application provides a solution that can effectively improve the efficiency, standardization, and accuracy of case handling by automatically generating interrogation outlines, providing real-time assistance in follow-up questioning, verifying the completeness of evidence, and automatically matching applicable legal provisions.

[0071] Therefore, this invention proposes a method for generating records based on a large model. It is understood that the computer product is equipped with a generation device for storing and executing the following method. The generation device can be implemented using a main controller, such as an MCU (Micro Controller Unit), a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), or a SOC (System-on-Chip).

[0072] In current technology, the recording of judicial cases has long relied on manual operation. Police officers need to manually conceive interrogation questions, assess the completeness of answers, and generate follow-up questions in real time. Traditional methods suffer from problems such as omissions in questioning, insufficient uncovering of key facts, and incomplete evidence chains. The process of drafting legal documents is cumbersome and prone to errors in standardization. Existing auxiliary tools mostly use fixed templates, which cannot adapt to complex case scenarios and lack in-depth semantic understanding and dynamic decision-making capabilities. For example, in theft cases involving multiple witnesses, police officers need to design the order of questioning different witnesses themselves, manually check for contradictions in testimony, and supplement missing information, which can easily lead to the omission of key evidence due to lack of experience.

[0073] To address the aforementioned issues, this paper proposes incorporating large-scale modeling technology into the transcript generation process, automatically generating an interrogation logic framework through case information parsing. First, the intelligent generation of interrogation outlines needs to be solved by constructing a questioning system using structured case elements. Second, dynamic verification of transcript information is required, identifying unanswered questions in real time and triggering follow-up questioning mechanisms. Finally, multi-source evidence materials need to be integrated to achieve automatic legal provision matching. Based on this, a multi-layered processing flow is designed, combining the semantic understanding capabilities of large-scale models with a legal knowledge graph to form a closed-loop intelligent transcript generation system.

[0074] Based on the above, referring to Figure 1 In one embodiment of the present invention, the method for generating transcripts based on a large model includes steps S100-S600, wherein:

[0075] S100. Obtain current case information;

[0076] S200: Generate several interrogation question points corresponding to the case information, and generate interrogation outline information corresponding to each interrogation question point, and display the interrogation outline information on the smart terminal;

[0077] S300, in response to the police officer's input of a first instruction on the smart terminal, determine the first transcript information corresponding to each of the interrogation outline information;

[0078] S400. Based on the first transcript information, determine the unanswered questions, generate corresponding follow-up questions, and in response to the police officer's input of a second instruction on the smart terminal, update the first transcript information to the target transcript information.

[0079] S500, obtain the forensic appraisal report and multiple pieces of supporting evidence;

[0080] S600. Based on the case information, target record information, judicial appraisal report, and multiple evidentiary materials, determine the corresponding applicable legal provisions and generate a legal document containing the applicable legal provisions.

[0081] Case information refers to the time, location, involved personnel, case type, and preliminary evidence materials of the case. Structured data can be extracted from police reports and on-site investigation reports using natural language processing technology. This case information serves as the foundational input for constructing the interrogation framework. Interrogation questioning points refer to a set of questions designed for the core elements of the case. These are generated through semantic parsing of case information using a large model. For example, in a theft case, questions about the source of the murder weapon and the whereabouts of stolen goods are automatically generated to ensure comprehensive coverage of key facts. The first instruction refers to the officer's operational command on a smart terminal to confirm the validity of the interrogation outline. This can be achieved through touchscreen clicks or voice commands and triggers the formal generation process of the transcript. Unanswered questions refer to logical contradictions or missing information in the transcript. These are identified through a pre-set integrity detection algorithm, such as conflicting testimonies or insufficient relevance of physical evidence, driving the execution of subsequent questioning mechanisms. A forensic expert opinion refers to a physical evidence examination report issued by a professional institution, which may include digital documents such as fingerprint comparison results and DNA testing data, serving as a key input source for verifying the chain of evidence.

[0082] The case information is analyzed using a large model to generate a hierarchical interrogation framework. The interrogation outline is organized according to dimensions such as the crime process, evidence correlation, and interpersonal relationships, and is displayed in separate areas on the smart terminal, showing the main questions and supplementary follow-up questions. Based on the actual questioning, the officer triggers a first command, and the generation device automatically extracts key information from the answers and generates a structured transcript. When contradictions or omissions are detected in the transcript, priority-ranked follow-up questions are automatically generated, and the officer updates the transcript by inputting supplementary information via a second command. Finally, the forensic report, evidence materials, and complete transcript are integrated, and a legal document containing applicable provisions is generated through legal knowledge graph matching. For example, in an assault case, Article 234 of the Criminal Law is automatically linked to generate a draft indictment.

[0083] Among its features, the system automatically generates key points for questioning based on case information to guide police officers in making statements; it extracts key points for statement collection from the current questions and answers to indicate whether any content is missing; and it automatically generates follow-up questions based on the current questions and answers for police officers to refer to.

[0084] The interrogation outline is automatically generated on the record page based on the cause of action and basic case information. You can select the outline to insert into the record.

[0085] Follow-up questions can be generated based on the answers in the transcript, and these follow-up questions can be selected and inserted into the transcript.

[0086] Based on the corresponding answers to the outline questions, you can check the completeness of the answers, display the answered and unanswered elements of the question, and insert follow-up questions for the unanswered elements.

[0087] By analyzing case information, interrogation records, forensic reports, and other documents, the system automatically generates a summary of the facts of the crime, recommends applicable legal provisions, and ultimately generates a legal document, which can be manually modified by police officers. After clicking "Save," the system will automatically conduct a document compliance review, checking for missing elements, typos, and issues related to the legal basis of the document.

[0088] The method for generating transcripts based on a large model in this embodiment can identify intent based on questions, and intelligent questioning can select different solutions according to different case types:

[0089] (1) Ask questions using the preset configuration.

[0090] (2) Based on the extraction of elements from the historical answers in the large model, intelligent questioning branching and slot filling are performed, and questions can be asked after completion.

[0091] Compared to existing technologies, traditional methods rely on manually designing questions, which carries the risk of logical inconsistencies or omissions. This embodiment, however, automatically generates a comprehensive interrogation framework covering all elements of the case using a large model, reducing human error. Fixed template tools cannot dynamically adjust questioning strategies; this embodiment employs a multi-turn dialogue model to generate follow-up questions in real time, adapting to complex case changes. Existing technologies require manual retrieval of legal provisions; this embodiment automatically matches applicable clauses in a knowledge graph based on the evidence chain integrity verification results, improving the efficiency of legal document drafting.

[0092] Through the above technical solutions, this embodiment realizes the intelligent generation and dynamic optimization of interrogation outlines, reducing the risk of omitting key facts; ensures the integrity and consistency of transcript information through a real-time follow-up questioning mechanism; and reduces the error rate of legal documents and improves the efficiency of judicial case handling by automatically matching legal provisions based on multi-source evidence materials.

[0093] This embodiment acquires current case information, generates several interrogation questioning points corresponding to the case information, and generates interrogation outlines corresponding to each interrogation questioning point. These interrogation outlines are then displayed on a smart terminal. In response to a police officer's input of a first command on the smart terminal, the first transcript information corresponding to each interrogation outline information is determined. Unanswered questions are identified based on the first transcript information, and corresponding follow-up questions are generated. In response to a police officer's input of a second command on the smart terminal, the first transcript information is updated to the target transcript information. Then, a forensic expert opinion and multiple pieces of evidentiary evidence are acquired. Finally, based on the case information, the target transcript information, the forensic expert opinion, and the multiple pieces of evidentiary evidence, the applicable legal provisions are determined, and a legal document containing the applicable legal provisions is generated. Thus, by automatically generating interrogation outlines, providing real-time assistance in follow-up questioning, verifying the completeness of evidence, and automatically matching applicable legal provisions, the efficiency, standardization, and accuracy of case handling can be effectively improved.

[0094] Optionally, refer to Figure 2 Another embodiment of the present invention provides a method for generating transcripts based on a large model, based on the above. Figure 1 The illustrated embodiment determines the applicable legal provisions based on the case information, target transcript information, forensic expert opinion, and multiple evidentiary materials, and generates a legal document containing the applicable legal provisions, including steps S610-S650, wherein:

[0095] S610. Analyze the legal elements of case information using a large model to extract a set of key factual features;

[0096] S620. Perform semantic association matching between the set of key fact features and the target transcript information to construct a case fact map;

[0097] S630. Verify the integrity of the evidence chain based on the aforementioned case fact map, judicial appraisal report, and / or multiple pieces of aforementioned evidentiary material, and determine the integrity of the evidence chain;

[0098] S640. Based on the completeness of the evidence chain, generate a list of applicable legal provisions recommendations by matching a preset legal provisions knowledge graph. The list of applicable legal provisions recommendations includes at least one applicable legal provision.

[0099] S650. Generate the legal document according to the recommended list of applicable legal provisions and the target record information.

[0100] Among these, legal element analysis refers to identifying core elements related to legal requirements from case information. This can be achieved using natural language processing technology combined with a legal knowledge base. Legal element analysis is used to accurately locate key facts affecting the application of law. Semantic association matching refers to establishing logical connections between different information sources by calculating the similarity of text vectors. This can be achieved using a semantic embedding model based on deep learning. Semantic association matching is used to construct a network of connections between case elements. Case fact graph refers to a data model that expresses case elements and their relationships in a graph structure. This can be achieved using knowledge graph construction technology. Case fact graph is used to visualize the core fact chain of a case. Evidence chain integrity verification refers to the quantitative evaluation of the logical closure between evidentiary materials. This can be achieved using a rule-based reasoning engine combined with a probabilistic model. Evidence chain integrity verification is used to detect loopholes in the evidence system. Pre-set legal provision knowledge graph refers to a relational database that stores legal provisions and their applicable conditions. This can be achieved using legal ontology modeling technology. Pre-set legal provision knowledge graph is used to establish a mapping relationship between legal provisions and case characteristics.

[0101] The process involves analyzing case information using a large-scale model to create structured legal elements. These elements are then semantically matched with the transcript information to generate a case fact graph containing nodes such as time, location, and actions. This graph is cross-validated with forensic reports, physical evidence, and other materials, and a logical reasoning engine detects any breaks or contradictions in the chain of evidence. A valid and complete chain of evidence triggers a matching mechanism within the legal knowledge graph, generating a recommendation list based on the similarity between case characteristics and the applicable conditions of the legal provisions. Finally, the process automatically generates legal documents containing accurate legal citations, achieving full automation from evidence analysis to document output.

[0102] Existing methods rely on manual comparison of legal provisions with case characteristics, which suffers from subjective judgment bias and inefficiency. Fixed-rule systems in existing technologies cannot handle the non-linear relationships between complex case elements, limiting the accuracy of legal provision matching. This embodiment dynamically models the association of case elements using a case fact graph, and combines this with a knowledge graph to achieve multi-dimensional matching of the applicable conditions of legal provisions, significantly improving the accuracy and comprehensiveness of legal provision citation.

[0103] This embodiment can automatically identify the core legal elements of a case and construct a visualized network of factual relationships, effectively detect logical loopholes in the chain of evidence, and intelligently match the most relevant legal provisions based on a knowledge graph. This technical solution solves the technical problems of traditional methods that rely on human experience for legal citation, which is prone to omissions or errors. At the same time, it improves the efficiency of legal document generation through structured data processing, ensuring the standardization and accuracy of legal application.

[0104] Optionally, refer to Figure 3 Another embodiment of the present invention provides a method for generating transcripts based on a large model, based on the above. Figure 2 The illustrated embodiment generates a recommended list of applicable legal provisions by matching a preset legal provision knowledge graph according to the completeness of the evidence chain, including steps S641-S644, wherein:

[0105] S641. Based on the assessment results of the integrity of the chain of evidence, calculate the applicability matching degree of each legal provision;

[0106] S642. Based on the matching degree, perform preliminary screening of legal provisions in the preset legal provision knowledge graph;

[0107] S643. Perform multi-dimensional feature weighted calculation on the filtered legal provisions to generate the priority ranking result of the legal provisions;

[0108] S644. Extract the top N legal provisions based on the priority sorting result to form a recommended list of applicable legal provisions, where N is an integer greater than or equal to 1.

[0109] The legal provision application matching degree refers to the degree of correlation between the legal provision and the factual characteristics of the case. This can be achieved by using a text similarity algorithm to calculate the semantic association value between the legal provision text and the case fact graph. This legal provision application matching degree is used to quantify the relevance between the legal provision and the case facts, avoiding omissions caused by subjective human judgment. The multi-dimensional feature weighted calculation refers to setting weight coefficients based on dimensions such as case type, legal provision timeliness, and frequency of judicial precedent citations. A linear weighted model can be used to score the selected legal provisions, improving the accuracy of legal provision matching through multi-dimensional feature fusion. The first N legal provisions refer to the dynamically adjusted number of recommendations based on the actual complexity of the case. This can be dynamically calculated using a preset threshold or the proportion of case elements. For example, N can be set to 3-5 in ordinary criminal cases, and 5-8 in cases involving multiple charges.

[0110] After the evidence chain integrity verification is completed, a semantic matching algorithm is first used to calculate the similarity between the case factual features and the legal provisions described in the knowledge graph, generating a matching score for each legal provision. Legal provisions with a matching score below a preset threshold are automatically filtered out, and the remaining legal provisions enter the weighted calculation stage. During the weighted calculation process, the generation device automatically loads the corresponding weight configuration according to the case type. For example, in theft cases, the weight of legal provisions related to "property value determination" is increased, and in injury cases, the priority of legal provisions related to "injury assessment standards" is increased. After weighted calculation, the legal provisions are arranged in descending order of total score, and finally, the top N legal provisions are selected according to preset rules to form a recommendation list.

[0111] Existing legal provision matching methods mostly rely on manual experience screening or simple keyword matching, which easily overlooks legal provisions with low relevance but practical applicability. This embodiment, through quantitative matching degree calculation and multi-dimensional feature fusion, can discover implicit connections between case facts and legal provisions. For example, it can automatically associate the factual feature of "multiple petty thefts" with the legal provision of "criminalization based on cumulative amount," solving the problem of cross-provisional application that is easily overlooked during manual retrieval.

[0112] This embodiment automates and refines the legal provision matching process, effectively resolving issues of omissions and mismatches in traditional methods. Through quantitative calculations and multi-dimensional evaluation, it ensures that the recommended legal provisions not only conform to the core facts of the case but also take into account complex situations in judicial practice, providing accurate legal basis for the generation of legal documents.

[0113] Optionally, refer to Figure 4 Another embodiment of the present invention provides a method for generating transcripts based on a large model, based on the above. Figure 1The illustrated embodiment generates several interrogation questioning points corresponding to the case information, and generates interrogation outline information corresponding to each interrogation questioning point, and displays the interrogation outline information on a smart terminal, including steps S210-S250, wherein:

[0114] S210. Determine the case type and cause of action characteristics corresponding to the case information;

[0115] S22. Match the corresponding interrogation template framework in the pre-set interrogation template library based on case type and cause of action features;

[0116] S230. Perform deep semantic analysis on the case information using a large model to generate a set of core interrogation elements corresponding to the interrogation template framework;

[0117] S240. Organize the core interrogation elements in a hierarchical manner according to the logical structure of the interrogation template framework to generate structured interrogation outline information.

[0118] S250. Based on the display layout specifications of the interrogation template framework, the interrogation outline information is dynamically and visually displayed in different areas on the smart terminal interface using an adaptive layout algorithm.

[0119] Case type and cause-of-fact characteristics refer to classification labels formed by analyzing basic case information. These can be achieved using natural language processing techniques to extract keywords from case description text and predict classification models. Case type and cause-of-fact characteristics guide the selection of subsequent interrogation frameworks. A pre-built interrogation template library is a database storing questioning frameworks corresponding to different case types. This library can use a tree structure to store standardized interrogation process templates for different causes of action, enabling rapid matching of basic questioning frameworks. Deep semantic parsing utilizes large models to understand the context of case materials. A pre-trained model based on a transformer architecture can be used for entity recognition and relation extraction, generating a core set of interrogation elements including timelines and motivational elements. Hierarchical organization involves structuring the extracted elements according to logical relationships. This hierarchical organization can use knowledge graph construction technology to associate elements with corresponding nodes in the template framework, forming an interrogation process with causal chains. Adaptive layout algorithms dynamically adjust the display method based on the characteristics of the terminal screen. Responsive web design technology combined with an element priority calculation module can be used to achieve dynamic partitioning and layout of primary and secondary information.

[0120] The process involves several steps. First, the input case information is parsed into case type tags and cause-of-fact feature vectors. These are then matched against a pre-set template library to retrieve the corresponding basic interrogation framework. The large model performs semantic analysis on the case details, extracting core elements such as the relationships between involved parties and the timeline of their actions. These elements are mapped to corresponding logical nodes within the template framework, forming a tree structure. The interrogation outline information is presented as a chain of key questions in the main display area, based on the strength of the association between elements. A secondary display area dynamically loads relevant evidence prompts. Screen layout parameters are calculated in real-time based on the terminal type; for example, a two-column layout is used on tablets to display the main question chain and auxiliary information, while a layered, collapsible menu is used on mobile devices.

[0121] Existing methods rely on manual experience to select fixed templates, making it difficult to dynamically adapt to the differences in case details, resulting in a lack of targeted questioning. This embodiment, however, generates a set of elements through deep analysis of a large model, and combines this with a template framework to achieve structured reorganization, ensuring that the interrogation outline maintains logical integrity while dynamically reflecting the characteristics of the case. Static layout methods in existing technologies cannot adapt to the display needs of different terminals, easily leading to information overload or the omission of key elements. The adaptive layout algorithm used in this embodiment can dynamically optimize the information presentation density according to the screen size, ensuring the visualization priority of core elements.

[0122] This embodiment addresses the issues of loose logic and missing elements in traditional interrogation outline generation, achieving dynamic adaptation between case characteristics and the interrogation framework. The generation of a structured set of interrogation elements avoids subjective biases from manual summarization, while the hierarchical organization ensures the rigor of causal relationships within the questioning chain. The dynamic visualization mechanism effectively enhances the operability of complex interrogation information across different terminal devices, reducing the risk of omitting key points during the interrogation process.

[0123] Optionally, refer to Figure 5 Furthermore, this invention provides a method for generating transcripts based on a large model, based on the above... Figure 4 The illustrated embodiment, based on the display layout specifications of the interrogation template framework, uses an adaptive layout algorithm to dynamically and visually display the interrogation outline information in different areas on the smart terminal interface, including steps S251-S255, wherein:

[0124] S251. Parse the display layout specifications of the interrogation template framework to obtain the display priority configuration and association configuration of each interrogation element;

[0125] S252 divides the interrogation outline information into a main display area and an auxiliary display area according to the display priority configuration, and establishes the association between the main display area and the auxiliary display area;

[0126] S253. Based on the association configuration and the screen characteristics of the smart terminal, dynamically calculate the optimal layout parameters and display style of the main display area and the auxiliary display area;

[0127] S254. Display the information on the smart terminal according to the optimal layout parameters and the display style;

[0128] S255. Adjust the display content of the main display area and the auxiliary display area in real time according to the progress of the interrogation.

[0129] The display priority configuration refers to the weighting of interrogation elements based on their importance in the case handling process. This can be achieved by mapping element type classification to case type associations, such as setting core factual elements as high priority to ensure key information is displayed first. The association configuration defines the logical dependencies between interrogation elements, which can be achieved through semantic graph construction technology, such as linking timelines with witness testimonies. This association configuration is used to assist in the dynamic linkage of content in the display area. The adaptive layout algorithm dynamically adjusts the display area ratio based on the terminal screen size and resolution. This can be achieved using a responsive layout engine combined with dynamic rendering technology, such as automatically adjusting the area distribution by calculating available screen space and element display density. The optimal layout parameters refer to the adaptation scheme of the position, size, and spacing of the main display area and the auxiliary display area. This can be achieved using a constrained optimization model to calculate the combination of layout parameters under different screen conditions, such as setting the main area to 70% in landscape mode.

[0130] The interrogation template framework's display layout specifications are parsed, and the generation device automatically identifies the priority and association rules of interrogation elements. The main display area is used to centrally display high-priority interrogation points, such as key time points in the case and relationships between involved persons, while the auxiliary display area displays supplementary materials or historical transcript fragments. During the operation of the smart terminal, the adaptive layout algorithm monitors parameters such as screen orientation and resolution in real time, dynamically adjusting the position distribution and content scaling ratio of the main and auxiliary areas. For example, when the terminal switches to portrait mode, the main area is adjusted to full-screen width, and the auxiliary area is placed at the bottom in the form of a collapsed label; when the interrogation process progresses to the evidence verification stage, the auxiliary area automatically displays corresponding physical evidence photos or expert report summaries.

[0131] This embodiment achieves hierarchical display and cross-terminal adaptation of interrogation elements through priority configuration and association rule definition, combined with an adaptive layout algorithm, while also supporting dynamic content updates driven by the interrogation stage. This embodiment solves the problems of chaotic interrogation information display and poor cross-device adaptability, realizing structured hierarchical display and dynamic linkage of the interrogation outline. The division of primary and secondary areas allows for focused presentation of core information, real-time linkage of related content reduces the burden of manual searching, the adaptive layout algorithm ensures display consistency across different terminal devices, and dynamic adjustments driven by the interrogation process improve the timeliness of information updates.

[0132] Optionally, refer to Figure 6 Another embodiment of the present invention provides a method for generating transcripts based on a large model, based on the above. Figure 1 The illustrated embodiment, in response to a police officer's input of a first instruction on the smart terminal, determines the first transcript information corresponding to each of the interrogation outline information, including steps S310-S330, wherein:

[0133] S310. Continuously collect the answers that police officers input on their smart terminals, which correspond to the information in each interrogation outline.

[0134] S320. Extract key information elements from the answer content using deep learning semantic analysis technology;

[0135] S330. Logically verify the extracted key information elements, and generate the first transcript information in a preset format from the verified key information elements.

[0136] Deep learning semantic analysis technology refers to the technique of feature extraction and intent recognition from natural language text based on neural network models. It can be implemented using bidirectional long short-term memory networks combined with attention mechanisms. Deep learning semantic analysis is used to automatically identify core elements related to the facts of the case from unstructured responses. Logical verification refers to the process of checking the internal consistency of extracted information. This can be implemented using conflict detection algorithms based on rule engines. Logical verification is used to discover contradictory statements or logical loopholes in the responses. Pre-formatted structured information generation refers to formatting the verified information according to legal document standards. This can be implemented using template-based text generation engines. Pre-formatted structured information generation ensures the standardization and traceability of transcript information.

[0137] During the interrogation, officers ask questions based on an interrogation outline displayed on a smart terminal and record the answers. The generation device continuously collects input content using real-time text stream processing technology and utilizes a pre-trained semantic analysis model to perform entity recognition and relationship extraction on the answers, such as automatically extracting case elements like time, location, and behavior. The verification module performs timeline comparison and factual correlation detection on the extracted elements; for example, it triggers anomaly markers when contradictory time descriptions of the same event are found. Verified elements are automatically populated into a preset transcript template, forming structured transcript data containing case element tags, question-and-answer correspondences, and logical verification identifiers.

[0138] Existing record-keeping methods rely on manual recording and post-processing, which carries the risk of information omissions and non-standard formatting. Current auxiliary tools only provide fixed-field input interfaces and cannot dynamically extract key elements from unstructured text. This embodiment, through deep semantic analysis and automated logical verification, achieves real-time parsing and structuring of free-text responses, overcoming the technical shortcomings of traditional methods such as incomplete information extraction and low record-generating efficiency.

[0139] This embodiment achieves automated element extraction and structured processing of interrogation responses, effectively reducing subjective errors in manual recording and improving the completeness and legal validity of the transcript information. Through a real-time logical verification mechanism, contradictions in the responses can be promptly identified, providing data support for subsequent questioning and preventing breaks in the chain of evidence due to missing information. The structured transcript data can be directly integrated with the legal document generation module, reducing the workload of manual transcription and format adjustment.

[0140] Optionally, refer to Figure 7 Another embodiment of the present invention provides a method for generating transcripts based on a large model, based on the above. Figure 1 The illustrated embodiment determines unanswered questions based on the first transcript information, generates corresponding follow-up questions, and updates the first transcript information to the target transcript information in response to the police officer's input of a second command on the smart terminal, including steps S410-S460, wherein:

[0141] S410. Analyze the first transcript information using an integrity detection algorithm to identify missing information points and contradictions;

[0142] S420. Based on the question-and-answer logical relationship, determine the key factors that need to be confirmed in the contradiction points;

[0143] S430. Generate a set of follow-up questions corresponding to the missing information points and the key factors through a multi-turn dialogue model;

[0144] S440. The set of follow-up questions is sorted according to importance and urgency using a sorting algorithm and displayed on the smart terminal.

[0145] S450, In response to the officer's input of a second instruction on the smart terminal, generate supplementary information corresponding to the set of follow-up questions;

[0146] S460. The supplementary information is fused with the first record information to generate the target record information.

[0147] Among them, the integrity detection algorithm refers to the algorithm used to detect missing information or logical contradictions in the transcript content. It can be implemented using dependency parsing and semantic role labeling in natural language processing technology. The integrity detection algorithm is used to automatically identify uncovered question points or contradictory statement fragments in the transcript. Question-answering logical relationship refers to the logical association rules between questions and answers. It can be implemented through a pre-built question-answering knowledge graph or logical reasoning model. The question-answering logical relationship is used to locate the core elements that need secondary verification in contradictory points. The multi-turn dialogue model refers to the generative model that can generate coherent follow-up questions based on context. It can be implemented using a sequence-to-sequence model based on attention mechanism. The multi-turn dialogue model is used to dynamically generate supplementary questions related to the facts of the case. The ranking algorithm refers to the calculation method that ranks the priority of questions according to preset rules. It can be implemented using a weighted scoring mechanism combined with urgency label classification. The ranking algorithm is used to optimize the display order of follow-up questions.

[0148] After generating the first transcript, the integrity detection algorithm performs semantic analysis on the transcript text, comparing it with the pre-set mandatory answers in the interrogation outline and the recorded content, marking unmentioned elements or contradictory statements. Based on a question-and-answer logical relationship model, causal reasoning is performed on contradictory points to identify key factors affecting the determination of facts. A multi-turn dialogue model generates a set of follow-up questions based on missing information points and key factors, such as generating verification questions for inconsistent timelines. A ranking algorithm dynamically adjusts the display priority of questions based on their impact on the integrity of the evidence chain and the urgency of the case. When the police officer inputs supplementary answers via a smart terminal, the generation device semantically aligns the new content with the original transcript and integrates the format, ultimately outputting a structurally complete and logically consistent target transcript.

[0149] Traditional methods rely on manual verification of transcript integrity, making it difficult to quickly identify hidden logical contradictions, and the generation of follow-up questions lacks specificity. This embodiment utilizes automated detection algorithms and a multi-turn dialogue model to systematically uncover unanswered questions and contradictions. Simultaneously, a priority ranking mechanism optimizes the guidance efficiency of the follow-up questioning process, avoiding evidence chain loopholes caused by human error. This embodiment solves the problems of high omission rates in traditional transcript creation due to manual judgment and strong randomness in follow-up question generation. It achieves automated verification of transcript integrity and targeted follow-up questioning guidance, effectively improving the accuracy of key fact discovery and the rigor of evidence chain construction.

[0150] Optionally, refer to Figure 8 Another embodiment of the present invention provides a method for generating transcripts based on a large model, based on the above. Figure 1 The embodiment shown further includes steps S700-S1000 in the method for generating transcripts based on a large model, wherein:

[0151] S700, Obtain legal templates and legal document formats;

[0152] S800. Examine the legal document according to the legal template and the legal document format, and determine the standardization level of the legal document;

[0153] S900. If the standardization level is lower than the preset standardization threshold, generate non-standard content annotations and corresponding modification suggestions;

[0154] S1000: In response to the police officer's modification command on the smart terminal, adjust the non-standard content of the legal document and generate a legal document that meets the standard.

[0155] Among them, legal templates refer to a pre-stored collection of legal document templates that conform to judicial norms. These can be implemented using publicly available standard document template libraries from judicial departments, providing format and content reference benchmarks for legal document generation. Legal document format refers to the standardized requirements for layout, chapter structure, and terminology that different types of legal documents should follow. This can be achieved by parsing judicial document format specification documents and constructing a structured format rule base, used for automated detection of document format deviations. The standardization level is a quantitative indicator of the degree to which the content and format of legal documents conform to preset standard norms. This can be achieved using a scoring algorithm based on a combination of rule engines and natural language processing technology, used to dynamically assess document compliance and trigger correction processes. Modification suggestions refer to correction guidance information generated for detected non-standard content. This can be achieved by comparing the differences between legal templates and the document under inspection and using large-scale model reasoning to generate correction schemes, used to assist police officers in quickly locating and correcting document errors.

[0156] This method automatically acquires legal templates and format rules through a large model, and performs multi-dimensional normative checks on the generated legal documents, such as checking the completeness of chapters, the format of legal citations, and the accuracy of terminology. When the normative level is detected to be below a preset threshold, the generation device automatically marks the problematic paragraphs or fields, and generates specific modification suggestions based on legal template difference analysis and semantic reasoning, such as supplementing missing chapters, adjusting the order of legal citations, and correcting terminology. Police officers can view the marked content and modification suggestions through a smart terminal, and choose to accept the suggestions or make manual adjustments according to the actual case. The generation device integrates the modified content with the original transcript information to regenerate a legal document that meets the normative requirements.

[0157] Existing methods rely on manual verification of legal documents for compliance, which is inefficient and prone to overlooking details. Existing auxiliary tools can only perform simple format checks based on fixed templates, failing to deeply analyze the compliance of content logic and legal elements. This embodiment, by integrating large-scale model semantic understanding and a rule engine, achieves dynamic and intelligent detection of legal document format and content. It can automatically identify hidden compliance defects and generate targeted modification suggestions, significantly improving the accuracy and efficiency of document correction.

[0158] This embodiment realizes automated verification and correction of the standardization of legal documents, solving the problems of low efficiency and error-proneness of traditional manual verification. At the same time, it overcomes the technical bottleneck that existing tools cannot deeply analyze the compliance of content, ensuring that the generated legal documents meet the requirements of judicial norms in terms of format, logic, and legal element citation, and effectively reducing the procedural risks caused by non-standard documents.

[0159] The present invention also proposes a generation apparatus, the generation apparatus comprising: a memory, a processor, and a large-model-based record generation program stored in the memory and executable on the processor, the large-model-based record generation program being configured to implement the large-model-based record generation method as described above.

[0160] It is worth noting that since the generation device of the present invention is based on the above-described method for generating transcripts based on large models, the embodiments of the generation device of the present invention include all the technical solutions of all embodiments of the above-described method for generating transcripts based on large models, and the technical effects achieved are exactly the same, so they will not be repeated here.

[0161] The memory refers to the physical device used to store program code and data, which can be implemented using solid-state drives or flash memory chips. Its function is to provide the processor with the storage space for the program instructions and case information required for operation. The processor is the hardware unit that executes program instructions to complete computational tasks. It can be implemented using a multi-core central processing unit (CPU). Its function is to realize case information parsing, interrogation outline generation, and legal provision matching by running a large-model-based transcript generation program. The large-model-based transcript generation program refers to a software module that includes natural language processing algorithms and knowledge graphs. It can be implemented by combining a pre-trained language model with legal domain fine-tuning. Its function is to perform semantic parsing of case information, generate structured interrogation outlines, and automatically match applicable legal provisions.

[0162] The generation device stores case information, interrogation templates, and a legal knowledge graph in its memory. When the processor executes the program, it first acquires case information and parses the case type, then calls a large model to generate key points for interrogation questions and corresponding interrogation outlines. Subsequently, it dynamically generates transcript information based on the instructions input by the police officer, and identifies missing information through an integrity detection algorithm to generate a set of follow-up questions. Finally, the device associates the case information, transcript content, and forensic reports, uses the legal knowledge graph to match applicable legal provisions, and automatically generates a compliant legal document.

[0163] In some specific implementations, the memory can integrate an encryption module to ensure data security, the processor can be equipped with an edge computing chip to improve real-time processing capabilities, and the program can support multi-threaded parallel processing to adapt to different case complexities. For example, during questioning, the device can use a voice recognition module to convert the conversation between the police officer and the suspect in real time and automatically fill it into the record information.

[0164] The present invention also proposes a computer product, which includes the generation apparatus as described in the above embodiments.

[0165] It is worth noting that since the computer product of the present invention is based on the above-described generating apparatus, the embodiments of the computer product of the present invention include all the technical solutions of all the embodiments of the above-described generating apparatus, and the technical effects achieved are exactly the same, so they will not be repeated here.

[0166] The generation device refers to a system entity integrating hardware and software, which can be implemented using embedded devices or server clusters. It generates transcripts through the collaborative execution of hardware resources and algorithm programs. The memory stores program code and case data, and can be implemented using solid-state drives or distributed databases, supporting high-concurrency read and write operations. The processor executes program instructions and can be implemented using multi-core CPUs or GPUs to meet the computational requirements of large-scale model reasoning. The transcript generation program based on the large model refers to the algorithm module deployed in the generation device. It can be implemented using a pre-trained language model combined with a legal knowledge graph, using natural language processing technology to parse case information and generate structured transcripts.

[0167] The generation device loads a transcript generation program based on a large model into its memory, which is then executed by a processor to drive the case information processing flow. During program execution, it first acquires case information and analyzes its legal elements, then dynamically generates an interrogation outline and key questions to assist officers in completing the transcript. In the transcript generation process, the program automatically links forensic opinions and evidence materials, verifies the completeness of the evidence chain, matches applicable legal provisions, and ultimately generates a compliant legal document. Through efficient scheduling of hardware resources and real-time reasoning of the algorithm model, the generation device achieves end-to-end automated processing from case information input to legal document output.

[0168] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0169] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0170] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0171] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for generating transcripts based on a large model, characterized in that, The method for generating transcripts based on a large model includes: Obtain current case information; Generate several interrogation question points corresponding to the case information, and generate interrogation outline information corresponding to each interrogation question point, and display the interrogation outline information on the smart terminal; In response to the police officer's input of a first instruction on the smart terminal, the first transcript information corresponding to each of the interrogation outline information is determined; Based on the first transcript information, unanswered questions are identified, corresponding follow-up questions are generated, and in response to the police officer's input of a second command on the smart terminal, the first transcript information is updated to the target transcript information. Obtain the forensic appraisal report and multiple pieces of supporting evidence; Based on the case information, target transcript information, forensic appraisal report, and multiple evidentiary materials, determine the corresponding applicable legal provisions and generate a legal document containing the applicable legal provisions; The process of determining the applicable legal provisions based on the case information, target transcript information, forensic expert opinion, and multiple evidentiary materials, and generating a legal document containing the applicable legal provisions, includes: The legal elements of case information are analyzed using a large model to extract a set of key factual features. The key fact feature set is semantically associated and matched with the target transcript information to construct a case fact map. Based on the aforementioned case fact map, forensic expert opinion, and / or multiple pieces of aforementioned evidentiary material, the integrity of the evidence chain is verified to determine the integrity of the evidence chain. Based on the completeness of the evidence chain, a list of applicable legal provisions is generated by matching a preset legal provisions knowledge graph. The list of applicable legal provisions includes at least one applicable legal provision. The legal document is generated according to the recommended list of applicable legal provisions and the target transcript information; The step of generating a recommended list of applicable legal provisions by matching a preset legal knowledge graph according to the completeness of the evidence chain includes: Based on the assessment results of the completeness of the chain of evidence, the applicability matching degree of each legal provision is calculated; Based on the matching degree, the legal provisions in the preset legal provisions knowledge graph are initially screened; The selected legal provisions are subjected to multi-dimensional feature weighting calculations to generate a priority ranking result for the legal provisions. Based on the priority ranking of the legal provisions, the top N legal provisions are extracted to form a recommended list of applicable legal provisions, where N is an integer greater than or equal to 1.

2. The transcript generation method based on a large model as described in claim 1, characterized in that, The process of generating several interrogation questioning points corresponding to the case information, generating interrogation outline information corresponding to each interrogation questioning point, and displaying the interrogation outline information on the smart terminal includes: Determine the case type and cause of action characteristics corresponding to the case information; Based on case type and cause of action features, match the corresponding interrogation template framework from the pre-set interrogation template library; By performing deep semantic analysis on the case information using a large model, a set of core interrogation elements corresponding to the interrogation template framework is generated. The core interrogation elements are organized hierarchically according to the logical structure of the interrogation template framework to generate structured interrogation outline information. Based on the display layout specifications of the aforementioned interrogation template framework, the interrogation outline information is dynamically and visually displayed in different areas on the smart terminal interface using an adaptive layout algorithm.

3. The transcript generation method based on a large model as described in claim 2, characterized in that, The display layout specification based on the interrogation template framework uses an adaptive layout algorithm to dynamically visualize the interrogation outline information in different areas on the smart terminal interface, including: The display layout specifications of the interrogation template framework are analyzed to obtain the display priority configuration and association configuration of each interrogation element; Based on the display priority configuration, the interrogation outline information is divided into a main display area and an auxiliary display area, and the association between the main display area and the auxiliary display area is established; Based on the configuration of association relationships and the screen characteristics of smart terminals, the optimal layout parameters and display styles of the main display area and the auxiliary display area are dynamically calculated; The display is shown on the smart terminal according to the optimal layout parameters and the display style; The display content of the main display area and the auxiliary display area are adjusted in real time according to the progress of the interrogation.

4. The transcript generation method based on a large model as described in claim 1, characterized in that, The step of responding to a police officer's input of a first instruction on the smart terminal to determine the first transcript information corresponding to each of the interrogation outline pieces of information includes: Continuously collect the answers that police officers input on their smart terminals, which correspond to the information in each interrogation outline; Key information elements are extracted from the answer content using deep learning semantic analysis technology; The extracted key information elements are logically verified, and the verified key information elements are used to generate the first transcript information in a preset format.

5. The transcript generation method based on a large model as described in claim 1, characterized in that, The step of determining unanswered questions based on the first transcript information, generating corresponding follow-up questions, and updating the first transcript information to the target transcript information in response to the police officer's input of a second command on the smart terminal includes: The first transcript information is analyzed using an integrity detection algorithm to identify missing information points and contradictions. Based on the question-and-answer logical relationship, the key factors that need to be confirmed in the points of contradiction are identified. A set of follow-up questions corresponding to the missing information points and the key factors is generated through a multi-turn dialogue model; The set of follow-up questions is prioritized according to importance and urgency using a sorting algorithm and then displayed on the smart terminal. In response to the police officer's input of a second command on the smart terminal, supplementary information corresponding to the set of follow-up questions is generated; The supplementary information is fused with the first transcript information to generate the target transcript information.

6. The transcript generation method based on a large model as described in claim 1, characterized in that, The large model-based transcript generation method also includes: Obtain legal templates and legal document formats; The legal document is examined according to the legal template and the legal document format to determine the level of standardization of the legal document; If the standardization level is lower than a preset standardization threshold, non-standard content annotations are generated and corresponding modification suggestions are generated; In response to the police officer's modification instructions on the smart terminal, the non-standard content of the legal document is adjusted to generate a legal document that meets the standard.

7. A generating apparatus, characterized in that, The generation apparatus includes: a memory, a processor, and a large-model-based transcript generation program stored in the memory and executable on the processor, the large-model-based transcript generation program being configured to implement the large-model-based transcript generation method as described in any one of claims 1 to 6.

8. A computer product, characterized in that, Includes the generating apparatus as described in claim 7.

Images