Mediation case intelligent analysis method and system based on AI large model

Abstract

The application belongs to the technical field of artificial intelligence, and specifically relates to a mediation case intelligent analysis method and system based on an AI large model, which comprises the following steps: inputting text materials, voice-to-text conversion texts and picture evidence of a mediation case into a preset multi-modal fusion large model; generating a high-dimensional vector representation; condensing a case core legal semantic feature vector; calculating a process failure risk index; judging whether the process failure risk index exceeds a preset risk threshold; generating a brand-new workflow topology structure and node task; calculating a process adaptability index; judging whether the process adaptability index is lower than a preset adaptability threshold; when the process adaptability index is lower than the preset adaptability threshold, generating a brand-new workflow topology structure and node task again; and when the process adaptability index is not lower than the preset adaptability threshold, ending the process. The application improves the accuracy and comprehensiveness of the case core semantic features.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of artificial intelligence, and specifically relates to a mediation case intelligent analysis method and system based on an AI large model. BACKGROUND

[0002] In the processing process, a case often involves multi-source heterogeneous data, including written materials submitted by parties, text converted from communication audio, and various picture evidence; these data collectively constitute a comprehensive description of the facts of the case, but their multi-modal and unstructured characteristics pose great challenges to traditional processing methods. Traditional mediation case management systems and methods mainly rely on fixed process templates driven by human experience and focus on processing structured text information; however, these methods have the following limitations:

[0003] Limitations of data processing: traditional systems have difficulty effectively integrating and processing heterogeneous data sources such as text, speech, and images; the lack of analysis capabilities for non-structured evidence such as speech and images leads to one-sided understanding of the facts of the case, making it difficult to build a complete evidence chain and easily ignoring key case information hidden in non-text information;

[0004] Limitations of semantic understanding: existing analysis methods are mostly limited to keyword matching or shallow text classification, and cannot perform deep legal semantic understanding on case materials; these methods are difficult to accurately capture complex relationships between multiple pieces of evidence, and cannot identify potential key evidence that can overturn the course of the case, leading to deviations in judging the core nature of the case;

[0005] Limitations of process management: current case management systems generally use pre-set, static workflow templates; such workflows lack the ability to perceive dynamic changes in case information, and when new important evidence appears, they cannot assess the applicability and compliance risks of existing processes, nor do they have the ability to dynamically adjust and optimize, relying instead on manual intervention, which is slow and prone to errors;

[0006] In recent years, artificial intelligence large model technology characterized by multi-modal fusion has made significant breakthroughs; by mapping data from different sources to a unified high-dimensional semantic space, this technology can achieve deep fusion and comprehensive understanding of case information, providing new possibilities for solving the above problems; multi-modal fusion large models can condense core legal semantic features representing the essence of the case from a global perspective, providing a solid foundation for dynamic assessment and decision-making;

[0007] In summary, in the prior art, the traditional case management system has significant deficiencies in handling multi-modal data, deep semantic understanding, and process dynamic adaptability; there is no research combining the deep semantic understanding ability of multi-modal fusion large model with the risk perception, dynamic reconstruction, and closed-loop correction mechanism of the workflow to build an intelligent and self-adaptive mediation case processing system that can actively adapt to changes in the case. SUMMARY

[0008] To solve the above technical problems, the present application discloses a mediation case intelligent analysis method and system based on AI large model, specifically, the technical solution of the present application is:

[0009] The mediation case intelligent analysis method based on AI large model comprises the following steps:

[0010] S1, input the text materials, voice transcription text and picture evidence of the mediation case into the preset multi-modal fusion large model;

[0011] S2, perform feature extraction on the text materials, voice transcription text and picture evidence through the multi-modal fusion large model to generate high-dimensional vector representation;

[0012] S3, perform deep fusion and analysis on the high-dimensional vector representation to condense the core legal semantic feature vector of the case;

[0013] S4, monitor the core legal semantic feature vector of the case in real time, combine the latest case legal semantic feature vector containing new evidence analyzed by the multi-modal fusion large model, and calculate the process failure risk index;

[0014] S5, judge whether the process failure risk index exceeds the preset risk threshold;

[0015] S6, when the process failure risk index exceeds the preset risk threshold, generate a new workflow topology structure and node task according to the latest case legal semantic feature vector and the preset legal compliance rule set;

[0016] S7, continuously monitor the execution of the new workflow topology structure and node task to calculate the process adaptability index;

[0017] S8, judge whether the process adaptability index is lower than the preset adaptability threshold;

[0018] S9, when the process adaptability index is lower than the preset adaptability threshold, generate a new workflow topology structure and node task again to form a closed-loop correction;

[0019] S10, when the process adaptability index is not lower than the preset adaptability threshold, end the process.

[0020] Preferably, S2 specifically comprises:

[0021] The multi-modal fusion large model extracts and embeds each modality data in the text material, voice transcription text, and picture evidence of the case, and generates a high-dimensional vector representation.

[0022] Preferably, S4 specifically comprises:

[0023] The cosine similarity between the latest case legal semantic feature vector containing new evidence and the case core legal semantic feature vector is calculated, and the process failure risk index is calculated in combination with the quantitative value of the impact of new evidence on the process.

[0024] Preferably, S6 specifically comprises:

[0025] The process reconstruction model constructed based on the multi-modal fusion large model automatically generates or recommends a brand new workflow topology structure and node task according to the latest case legal semantic feature vector and the preset legal compliance rule set.

[0026] Preferably, S7 specifically comprises:

[0027] The process execution effect quantitative value evaluates the matching degree with the latest case semantics by monitoring the intermediate results such as task completion rate and legal risk report, and the process adaptability index is calculated in combination with the process execution effect quantitative value and the risk index that may be generated in the process of reconstructing the brand new workflow topology structure and node task.

[0028] Preferably, S9 specifically comprises:

[0029] The sensing-reconstruction cycle is triggered again to fine-tune or reconstruct the process until the process adaptability index reaches the preset adaptability threshold.

[0030] The mediation case intelligent analysis system based on AI large model comprises:

[0031] The data input module is used to input the text material, voice transcription text, and picture evidence of the case into the multi-modal fusion large model;

[0032] The feature extraction module is used to extract features from the text material, voice transcription text, and picture evidence through the multi-modal fusion large model, and generate a high-dimensional vector representation;

[0033] The semantic condensation module is used to deeply fuse and analyze the high-dimensional vector representation, and condense the case core legal semantic feature vector;

[0034] The risk assessment module is used to monitor the case core legal semantic feature vector, and calculate the process failure risk index in combination with the latest case legal semantic feature vector containing new evidence analyzed by the multi-modal fusion large model;

[0035] A process reconfiguration module is configured to generate a brand-new workflow topology structure and node task according to the latest case legal semantic feature vector and the preset legal compliance rule set when the process failure risk index exceeds the preset risk threshold value.

[0036] An effect evaluation module is configured to continuously monitor the execution of the brand-new workflow topology structure and node task, and calculate a process adaptability index.

[0037] A closed-loop correction module is configured to generate a brand-new workflow topology structure and node task again to form a closed-loop correction when the process adaptability index is lower than a preset adaptability threshold value.

[0038] Preferably, the mediation case intelligent analysis system based on an AI large model comprises:

[0039] A processor, a memory, and a communication interface, the memory stores program code, and the processor is configured to execute the program code.

[0040] Compared with the prior art, the present application has the following beneficial effects:

[0041] 1. The present application realizes comprehensive collection and deep semantic understanding of multi-source heterogeneous data of a case by inputting the text materials, voice transcription text and picture evidence of the case into a multi-modal fusion large model. The method first extracts and embeds each type of modal data independently, and then performs deep fusion analysis to condense the core legal semantic feature vector of the case. This separation extraction and deep fusion strategy retains the unique semantic dimensions of various types of data, can capture key non-text information in oral statements, tone and image evidence in addition to written materials, improves the accuracy and comprehensiveness of the core semantic features of the case, and enhances the recognition ability of the system for revolutionary evidence hidden in non-text data.

[0042] 2.The application establishes a dynamic risk early warning and process reconstruction mechanism; the system monitors the core legal semantic feature vector of the case in real time. When new evidence appears, the semantic similarity between the latest case legal semantic feature vector containing the new evidence and the original core vector is calculated, and the quantized value of the influence of the new evidence on the process is combined to calculate a process failure risk index. This risk measurement method upgrades the fuzzy difference judgment to a comprehensive evaluation of the importance and magnitude of the change, and can accurately identify the key case changes that truly need to trigger process reconstruction. When the risk exceeds the threshold, the process reconstruction model based on the large model will automatically generate a new workflow topology and node task according to the latest case profile and the preset legal compliance rules. This converts the previous process design work relying on human experience into automatic generation driven by AI models, significantly improving the system's response speed to case mutations and decision-making quality, and upgrading process management from rigid task execution to dynamic self-reconstruction;

[0043] 3.The application realizes continuous self-optimization of the process through closed-loop correction; after the execution of the new workflow, the system will continuously monitor the task completion and legal risks, and calculate a process adaptability index based on the risk index of the process reconstruction itself, to quantitatively evaluate the actual effectiveness of the new process. This index comprehensively weighs the execution benefits and potential costs of the new process; when the adaptability is lower than the preset threshold, the system will trigger the perception and reconstruction cycle again to fine-tune or reconstruct the workflow, and learn from the ineffective reconstruction until the process adaptability meets the requirements. This iterative optimization closed-loop mechanism ensures that the system does not stay at a suboptimal solution, and ensures that no matter how complex the case is, the system can eventually converge to an efficient, compliant and practically proven optimal workflow through self-evolution, significantly improving the mediation success rate and processing quality of difficult cases. BRIEF DESCRIPTION OF DRAWINGS

[0044] The application will be further explained below in conjunction with the accompanying drawings and examples:

[0045] Figure 1 is a process flow diagram of the method of the application.

[0046] Figure 2 is a process flow diagram of the system of the application. DETAILED DESCRIPTION

[0047] To make the purpose, technical scheme and advantages of the application clearer, the application will be further described in detail below in conjunction with specific examples.

[0048] Example 1:

[0049] Please refer to Figure 1, the mediation case intelligent analysis method based on the AI large model comprises the following steps:

[0050] S1, input the text materials, voice transcription text and picture evidence of the mediation case to a preset multi-modal fusion large model;

[0051] S2, the multi-modal fusion large model is used for feature extraction of the text materials, voice transcription text and picture evidence, and high-dimensional vector representation is generated;

[0052] S3, the high-dimensional vector representation is deeply fused and analyzed, and the case core legal semantic feature vector is condensed;

[0053] S4, the case core legal semantic feature vector is monitored in real time, the latest case legal semantic feature vector containing new evidence analyzed by the multi-modal fusion large model is combined, and the process failure risk index is calculated;

[0054] S5, whether the process failure risk index exceeds the preset risk threshold value is judged;

[0055] S6, when the process failure risk index exceeds the preset risk threshold value, a new workflow topology structure and node task are generated according to the latest case legal semantic feature vector and the preset legal compliance rule set;

[0056] S7, the execution of the new workflow topology structure and node task is continuously monitored, and the process adaptability index is calculated;

[0057] S8, whether the process adaptability index is lower than the preset adaptability threshold value is judged;

[0058] S9, when the process adaptability index is lower than the preset adaptability threshold value, a new workflow topology structure and node task are generated again to form a closed loop correction;

[0059] S10, when the process adaptability index is not lower than the preset adaptability threshold value, the process is ended.

[0060] The embodiment of the application provides a mediation case intelligent analysis method based on an AI large model, the method constitutes a complete and self-consistent technical closed loop, and the specific implementation manner is as follows:

[0061] S1, input the text materials, voice transcription text and picture evidence of the mediation case to a preset multi-modal fusion large model;

[0062] The purpose of this step is to achieve comprehensive collection of multi-source heterogeneous data in the case. In a specific mediation case scenario, written materials refer to structured and unstructured texts such as complaints, answers, and lists of evidence submitted by the parties, which serve to provide the basic facts and legal claims of the case. Speech-to-text refers to text generated by automatic speech recognition (ASR) of court recordings and telephone communication recordings during the mediation process, which serves to capture oral statements, tone, and key information beyond written materials. Image evidence refers to image data such as photos, scans, and screenshots, such as scanned copies of contracts and photographs of physical evidence, which serve to provide intuitive, non-textual evidence support. These three types of data together constitute a holographic description of the case. Inputting this data into a pre-set multimodal fusion model is the foundation for subsequent deep semantic analysis. The pre-set multimodal fusion model refers to a deep learning model pre-trained with massive amounts of legal data, capable of understanding and processing multiple data modalities, and its core function is to serve as a unified semantic understanding engine for all subsequent intelligent analyses.

[0063] S2. Through a multimodal fusion model, features are extracted from textual materials, speech-to-text transcriptions, and image evidence to generate high-dimensional vector representations.

[0064] The purpose of this step is to transform unstructured multimodal data into a machine-understandable mathematical representation. The large multimodal fusion model integrates encoders for different data types; for example, it uses models like BERT as text encoders to process written materials and speech-to-text transcription, and models like CNN or ViT as image encoders to process image evidence. These encoders map the raw data into a high-dimensional vector space, generating a high-dimensional vector representation. Specifically, written materials are converted into vectors. The speech-to-text is converted into a vector. The visual evidence was converted into vectors. These vectors are not only a compact mathematical representation of the original data, but more importantly, their positions and orientations in the vector space contain deep semantic information about the data.

[0065] S3. Deeply integrate and analyze the high-dimensional vector representation to extract the core legal semantic feature vector of the case;

[0066] The purpose of this step is to extract the core semantics that have a decisive impact on the legal characterization of the case from the scattered multimodal information; the multimodal fusion big model uses attention mechanisms or other fusion strategies to represent high-dimensional vectors from different modalities ( ) to perform weighted fusion and deep analysis; this process is not simply a vector splicing, but simulates the process of human legal experts analyzing different sources of evidence, identifying and amplifying key information, and suppressing redundant or contradictory information; finally, the model condenses and outputs a single case core legal semantic feature vector ; the case core legal semantic feature vector ( ) refers to a high-dimensional vector that can comprehensively and accurately represent the legal attributes of the current case, and its role is as a benchmark and starting point for the entire subsequent process analysis and decision-making; its calculation process can be formally represented as: , wherein, represents the fusion and analysis function of the multi-modal fusion large model, and the input is the high-dimensional vector representation of each modality generated by S2; this vector can capture potential disruptive evidence or related case information that traditional methods cannot find;

[0067] S4, real-time monitoring of the case core legal semantic feature vector, combining the latest case legal semantic feature vector containing new evidence analyzed by the multi-modal fusion large model, to calculate the process failure risk index;

[0068] The purpose of this step is to quantify the impact of new evidence on the stability of the existing mediation process; in the mediation process, any newly submitted evidence will be input into the system and analyzed by the multi-modal fusion large model in real time to generate a latest case legal semantic feature vector containing new evidence; the system calculates the process failure risk index by comparing the difference between the new vector and the original case core legal semantic feature vector ; the process failure risk index ( ) refers to an index for measuring the risk of the current mediation process becoming unsuitable or illegal due to significant changes in the case; its role is as a decision basis for triggering process reconstruction;

[0069] S5, judge whether the process failure risk index exceeds the preset risk threshold;

[0070] This step is a decision triggering link; the system compares the process failure risk index calculated in the previous step with a preset risk threshold ; the preset risk threshold ( ) refers to a critical value set according to historical case data and legal expert experience; to clarify its setting logic, this threshold can be calculated by the risk index The 99% percentile of the probability distribution of the values is taken to determine the value, which serves to balance the need for the system to respond to significant changes in the case and to avoid frequent reconstruction of the process due to minor fluctuations.

[0071] S6、When the process failure risk index exceeds the preset risk threshold, a new workflow topology structure and node task are generated based on the latest case legal semantic feature vector and the preset legal compliance rule set;

[0072] When , the system determines that the current process has a legal or effective crisis and must be reconstructed; at this time, the system calls a process reconstruction model, which takes the latest case legal semantic feature vector as input, as it represents the most comprehensive state of the case; at the same time, the model must also follow a preset legal compliance rule set ); the preset legal compliance rule set refers to a rule library that encodes relevant laws, judicial interpretations, and procedural requirements, such as the provisions of the Civil Procedure Law regarding mediation procedures, which serves to ensure that the newly generated workflow is legally compliant; the process reconstruction model is based on an understanding of and is constrained by , and outputs a new workflow topology structure and node task ;

[0073] For example, if new evidence indicates that the case involves corporate bankruptcy, the new workflow may automatically add necessary nodes such as notifying the bankruptcy administrator and reviewing creditor claims, and the new workflow topology structure and node task are then loaded into the process execution engine of the system, which is responsible for parsing the workflow and scheduling resources to execute the node tasks it contains.

[0074] S7、Continuously monitor the execution of the new workflow topology structure and node task, and calculate the process adaptability index;

[0075] The purpose of this step is to evaluate the actual running effect of the newly generated workflow and form a feedback mechanism; during the execution of the new workflow topology structure and node task , the system continuously monitors its key performance indicators, such as the completion rate of each node and whether there are legal risk warnings, and calculates a process adaptability index based on these monitoring data; the process adaptability index is a comprehensive indicator that quantitatively assesses the matching degree and execution effectiveness of the new workflow with the current complex case, and its value range is usually between [0, 1], which serves to determine whether the reconstructed process has achieved the expected effect;

[0076] S8, judge whether the process adaptability index is lower than the preset adaptability threshold value;

[0077] Similar to risk judgment, this step is a decision trigger link of closed-loop correction; the system compares the process adaptability index calculated in the last step with a preset adaptability threshold value ; the preset adaptability threshold value refers to a critical value representing the minimum acceptable process efficiency; the setting basis is to set a reasonable value that can guarantee the basic effectiveness of the process and avoid endless fine-tuning due to the pursuit of perfection, so as to maintain the stability of the system, by analyzing the process adaptability index distribution of historical successful mediation cases;

[0078] S9, when the process adaptability index is lower than the preset adaptability threshold value, a brand new workflow topology structure and node task are generated again to form a closed-loop correction;

[0079] When , it indicates that the reconstructed process is not effective in actual operation and cannot effectively solve the problem; the system will trigger the perception-reconstruction cycle again, that is, repeat the operation of step S6 and the following steps to further fine-tune or completely reconstruct the workflow; this process is iterated until the adaptability index of the newly generated workflow meets the requirements, thereby forming a closed-loop correction;

[0080] S10, when the process adaptability index is not lower than the preset adaptability threshold value, end the process;

[0081] When , the system considers that the current workflow is effective, adaptive, and stable; at this time, the correction cycle terminates, and the system will continue to execute the current workflow until the mediation is completed;

[0082] Through the above steps, the present application constructs a complete intelligent processing process from multi-modal data understanding, risk perception, dynamic reconstruction to closed-loop correction; it makes the system not a rigid task executor, but a self-adaptive process management core that can actively perceive changes in case, warn process failure risks, and dynamically adjust its structure to adapt to new legal realities, greatly improving the intelligent level, compliance, and agility in dealing with complex and unexpected situations in mediation cases.

[0083] Embodiment 2:

[0084] S2 specifically includes:

[0085] The multi-modal fusion large model extracts and embeds features from each modal data in the case's text materials, voice transcribed text, and picture evidence, generating high-dimensional vector representations.

[0086] The embodiment is further limited on the technical details of step S2 based on example 1; the multi-modal fusion large model extracts and embeds each modal data in the case's text materials, voice transcription text and picture evidence to generate high-dimensional vector representation; specifically, this process ensures that the unique information of each data source is fully preserved and deeply mined; for example, for text materials, the model not only focuses on keywords, but also understands the legal logic through context analysis; for voice transcription text, the model can capture the semantic information implied by hesitation, emphasis and other tone words; for picture evidence, the model can identify the authenticity of seals, traces of tampering of key content and other non-text information; by independently generating the optimal vector representation for each modal data and then performing subsequent fusion, semantic loss in the initial information conversion can be minimized, and a high-quality case core legal semantic feature vector can be extracted Lay a solid foundation;

[0087] This separation and extraction, deep fusion strategy brings more refined feature extraction effect; compared with the way of simply text processing after all data, the invention can preserve the unique semantic dimension of each modal data, thereby improving the accuracy and comprehensiveness of the semantic feature vector, and enhancing the system's ability to identify revolutionary evidence hidden in non-text data.

[0088] Example 3:

[0089] S4 specifically includes:

[0090] Calculate the cosine similarity between the latest case legal semantic feature vector containing new evidence and the case core legal semantic feature vector, and combine the quantified value of the impact of new evidence on the process to calculate the process failure risk index.

[0091] The embodiment is further limited on the calculation method of the process failure risk index in step S4 based on example 1; calculate the cosine similarity between the latest case legal semantic feature vector containing new evidence and the case core legal semantic feature vector, and combine the quantified value of the impact of new evidence on the process to calculate the process failure risk index; the calculation follows the following formula: ;

[0092] Wherein: : process failure risk index, its technical meaning is the quantitative representation of the deviation of the case change introduced by new evidence from the original case basis, it is a dimensionless scalar value, and its source is the calculation result of the formula;

[0093] : the quantified value of the process impact of new evidence, whose technical meaning is the importance or subversive level of the evidence itself, is a normalized scalar value in the interval [0, 1]; its source is based on pre-set rules or AI model evaluation; for further clarification, the pre-set rules are a knowledge base formed by mapping specific evidence types to high impact force values according to the experience of legal experts;

[0094] : the cosine similarity between two vectors, whose technical meaning is the correlation degree of new and old case facts in the semantic space, whose value range is [0, 1], which is obtained by calculating the dot product of vectors and and normalizing it; the closer the value is to 1, the more similar the new and old case facts are; the closer to 0, the greater the difference is;

[0095] : the latest case legal semantic feature vector containing new evidence analyzed by the AI large model, whose source is the first half of step S4, which is the real-time processing result of the model on new evidence;

[0096] : the core legal semantic feature vector of the current process, whose source is step S3;

[0097] : a very small positive number, which prevents the denominator from being zero and ensures the stability of the calculation, whose source is a constant preset by the system, such as ;

[0098] Based on the above definitions, the technical motivation of this formula is that how much the flow is different depends on the importance of the new evidence itself; a very important new evidence may cause a huge process risk even if the overall case changes little;

[0099] The application of this formula brings more accurate risk measurement capability; it upgrades the risk assessment from fuzzy difference judgment to two-dimensional evaluation of the importance of change direction and the magnitude of change, so that it can more accurately identify those high-value case changes that really need to trigger process reconstruction, avoiding excessive reaction to non-critical changes, and improving the decision-making quality and operational efficiency of the system.

[0100] Embodiment 4:

[0101] S6 specifically includes:

[0102] The process reconstruction model constructed based on the multi-modal fusion large model automatically generates or recommends a completely new workflow topology and node task according to the latest case legal semantic feature vector and the pre-set legal compliance rule set.

[0103] The embodiment is based on example 1, and the way of generating a brand-new workflow in step S6 is specified; based on the process reconstruction model constructed by the multi-modal fusion large model, a brand-new workflow topology structure and node task are automatically generated or recommended according to the latest case legal semantic feature vector and the preset legal compliance rule set; the workflow can be formally expressed as: ;

[0104] Among them: : the brand-new workflow topology structure and node task generated by the AI large model according to the new situation, the data type of which can be a directed acyclic graph or a structured task list, and the source is the output of the model;

[0105] : the process reconstruction model constructed based on the AI large model, which acts as the core generation engine for generating new processes; in order to clarify its construction process, the model is trained in a labeled data set containing a large number of historical cases, wherein the variables of the data set are not input variables during model running, but data pairs composed of 'final case feature vector' of historical cases and 'optimal process template' corresponding to the case; the model learns the mapping relationship from the case to the process strategy by learning these data pairs;

[0106] : the latest case legal semantic feature vector, which provides a complete and accurate portrait of the current case for the model, and is the input during model running;

[0107] : the preset legal compliance rule set, which constrains the output space of the model to ensure that any generated process does not violate mandatory legal provisions;

[0108] This embodiment realizes the intelligence and automation of process reconstruction; it changes the process design work relying on artificial experience and manual adjustment in the past into automatic generation by the AI model based on the deep understanding of the nature of the case; this not only greatly improves the response speed, but more importantly, it can generate solutions that surpass human experience and are more optimal in law, so that the process engine is upgraded from an unsmart executor to an intelligent process core that can dynamically reconstruct itself.

[0109] Example 5:

[0110] S7 specifically includes:

[0111] The process execution effect quantitative value is evaluated by monitoring the task completion rate, legal risk report and other intermediate results to evaluate the matching degree with the latest case semantics, and the process adaptability index is calculated in combination with the process execution effect quantitative value and the risk index that may be generated in the process of reconstructing the brand-new workflow topology structure and node task.

[0112] This implementation method is based on Example 1, and adjusts the process adaptability index in step S7. The calculation method has been specified; the quantitative value of process execution effect is evaluated by monitoring intermediate results such as task completion rate and legal risk reports to assess the degree of matching with the semantics of the latest cases; the process adaptability index is calculated by combining the quantitative value of process execution effect and the risk index that may be generated during the reconstruction of the new workflow topology and node tasks; the calculation follows the following formula: ;

[0113] in: The process adaptability index, technically speaking, is a comprehensive trade-off between the benefits and costs of a new process. It is a scalar value normalized to the [0,1] interval and is derived from the calculation results of this formula.

[0114] The quantitative value of process execution effect is technically defined as the degree to which the new process matches the expected goals in actual operation. It is a scalar value normalized to the range of [0,1]. It is derived by continuously monitoring the execution status of each node in the process, the feedback from mediators, and whether any new legal risk warnings are triggered, and is calculated by a preset weighting function.

[0115] The risk index, which represents the potential risks arising from the restructuring of the new process, is technically an estimate of the potential problems caused by the complexity and uncertainty of the new process itself. It is a scalar value normalized to the [0,1] interval; its source is the new process. After generation and before formal execution, another AI evaluation model simulates and extrapolates the process to assess potential risks such as task allocation conflicts and omissions in legal clauses. This AI evaluation model itself is obtained through supervised learning of the process structure characteristics and actual failure mode data of historical reconstruction failure cases.

[0116] This formula ensures that the evaluation of the new process is comprehensive, taking into account both how many old problems it solves. We also need to consider whether it introduces new problems. This allows for closed-loop control of process quality.

[0117] This implementation method endows the system with the ability to provide early warning of system risks and to self-evolve; by conducting a simulation assessment of the risks of the new process before reconstruction ( The system can provide managers with decision-making insights, avoiding the hasty adoption of a new process with potential flaws; and through comprehensive evaluation of execution effectiveness ( ), the system can determine whether the reconstruction is successful and, on this basis, proceed to the next round of optimization iteration; this makes the system no longer a simple automation tool, but an intelligent platform that can master unknown complexity and constantly improve itself when handling difficult cases.

[0118] Embodiment 6:

[0119] S9 specifically includes:

[0120] The perception-reconstruction cycle is triggered again to fine-tune or re-reconstruct the process until the process adaptability index reaches the preset adaptability threshold.

[0121] This embodiment is a further clarification of the closed-loop correction mechanism in step S9 of embodiment 1; the perception-reconstruction cycle is triggered again to fine-tune or re-reconstruct the process until the process adaptability index reaches the preset adaptability threshold; this means that when , the system does not simply fall back to the original process, but takes the current, ineffective new process and its corresponding adaptability index as new input information to start the process reconstruction model again; the model learns from this unsuccessful reconstruction and avoids similar problems when generating the process next time; this cycle continues, which may be a completely new reconstruction or just a partial fine-tuning of the existing process nodes, until the adaptability index of the final version of the workflow satisfies ;

[0122] This embodiment ensures the convergence and final effect of process optimization; it establishes a closed loop of continuous feedback and iterative optimization to ensure that the system does not stay at any suboptimal solution; this correction mechanism ensures that no matter how complex and variable the case is, the system can always evolve itself and finally converge to an efficient, compliant and practically tested optimal workflow, significantly improving the mediation success rate and handling quality of difficult cases.

[0123] Embodiment 7:

[0124] Please refer to Figure 2 , the mediation case intelligent analysis system based on AI large model includes:

[0125] A data input module for inputting the text materials, voice-to-text and picture evidence of a case into a multi-modal fusion large model;

[0126] A feature extraction module for extracting features from the text materials, voice-to-text and picture evidence through the multi-modal fusion large model to generate high-dimensional vector representations;​

[0127] a semantic condensing module, configured to perform deep fusion and analysis on the high-dimensional vector representation, and condense a core legal semantic feature vector of the case;

[0128] a risk assessment module, configured to monitor the core legal semantic feature vector of the case, combine a latest legal semantic feature vector of the case containing new evidence analyzed by the multi-modal fusion large model, and calculate a process failure risk index;

[0129] a process reconstruction module, configured to, when the process failure risk index exceeds a preset risk threshold, generate a new workflow topology structure and node task according to the latest legal semantic feature vector of the case and a preset legal compliance rule set;

[0130] an effect evaluation module, configured to continuously monitor an execution condition of the new workflow topology structure and node task, and calculate a process adaptability index;

[0131] a closed-loop correction module, configured to, when the process adaptability index is lower than a preset adaptability threshold, generate a new workflow topology structure and node task again to form a closed-loop correction.

[0132] The application further provides an AI large model-based mediation case intelligent analysis system, which is a hardware carrier and function implementation of the method; the system comprises:

[0133] a data input module, configured to collect all information of a case, and in the embodiment, the module is responsible for receiving and preprocessing text materials uploaded by a user, speech-to-text conversion and picture evidence, and transmitting the multi-modal data to subsequent modules;

[0134] a feature extraction module, configured to convert original data into a semantic vector understandable by a machine, and in the embodiment, the module is internally provided with various encoders of a multi-modal fusion large model, and is configured to perform deep feature extraction on different types of data, and generate a high-dimensional vector representation;

[0135] a semantic condensing module, configured to form a unified and core cognition of a case, and in the embodiment, the module utilizes a fusion mechanism of a multi-modal fusion large model to perform deep fusion and analysis on the high-dimensional vector representation, and condenses a core legal semantic feature vector of the case ;

[0136] a risk assessment module, configured to perceive a process risk caused by a change in a case in real time, and in the embodiment, the module monitors a core legal semantic feature vector of the case , and when new evidence is input, combines a new vector analyzed, and calculates a process failure risk index ;

[0137] A process reconstruction module is configured to dynamically generate a mediation process that adapts to new case conditions. In this embodiment, when the process failure risk index exceeds a preset risk threshold, the internal process reconstruction model is called to generate a new workflow topology and node task according to the latest case legal semantic feature vector and a preset legal compliance rule set.

[0138] An effect evaluation module is configured to verify the actual effectiveness of the new process. In this embodiment, it continuously monitors the execution of the new workflow and calculates the process adaptability index according to the formula.

[0139] A closed-loop correction module is configured to realize self-optimization and evolution of the system. In this embodiment, when the process adaptability index is lower than a preset adaptability threshold, the process reconstruction module is instructed to generate a new workflow again to form a closed-loop correction.

[0140] A model verification module is configured to ensure the robustness of the system. In this embodiment, this module is responsible for stress testing the core model before it is formally deployed, using an independent test set containing historical difficult cases, edge cases, and simulated adversarial samples. It verifies whether the behavior of the model when processing extreme input values meets the logical expectations, and continuously monitors the output of the online model to prevent performance degradation caused by data drift or malicious input, thereby ensuring the safe and stable operation of the entire system in various complex or adversarial scenarios.

[0141] Through the organic combination and collaborative work of the above modules, the system physically realizes an end-to-end intelligent case processing capability. It deeply couples data collection, semantic understanding, risk assessment, decision generation, and effect feedback functions to form an efficient whole, providing a complete system-level solution for realizing intelligent and adaptive management of mediation cases throughout the entire process.

[0142] Embodiment 8:

[0143] The mediation case intelligent analysis system based on an AI large model includes:

[0144] A processor, a memory, and a communication interface, the memory storing program code, and the processor being configured to execute the program code. This embodiment is a specific example of the hardware implementation of the system based on Embodiment 7. The system can be a server, a computer, or a distributed computing cluster, and its core hardware at least includes a processor, a memory, and a communication interface.

[0145] ​​​​​Among them, the processor is the computing core of the system, responsible for executing program instructions, running the algorithm logic of each module, such as executing the inference calculation of the multi-modal fusion large model, calculating the risk index and adaptability index, etc.; the memory is used to store program codes, that is, software programs that solidify the above method processes and module functions, and is also used to temporarily store case data, intermediate calculation results and model parameters, etc.; the communication interface is used for data interaction with external devices, realizing the input and output of data; the processor is used to execute program codes, load and run instructions in the memory, thereby driving the entire system to realize all the functions described in embodiments 1 to 7.

[0146] The present embodiment clarifies the implementability of the technical solution of the present application, provides a specific hardware architecture, and ensures that all the above-mentioned methods and module functions have their physical execution carriers, so that the technical solution of the present application can be converted from theory to practical application products, and has the foundation of commercialization and engineering.

[0147] The above is only a preferred embodiment of the present application, and is not intended to limit the protection scope of the present application; any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0148] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit the protection scope of the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. An AI large model-based mediation case intelligent analysis method, characterized in that, The method comprises the following steps: S1, inputting the text materials of the mediation case, the speech transcription text, and the picture evidence into a preset multi-modal fusion large model; S2, extracting features of the text materials, the speech transcription text, and the picture evidence through the multi-modal fusion large model to generate high-dimensional vector representations; S3, deeply fusing and analyzing the high-dimensional vector representations to condense a case core legal semantic feature vector; S4, monitoring the case core legal semantic feature vector in real time, combining the latest case legal semantic feature vector containing new evidence analyzed by the multi-modal fusion large model to calculate a process failure risk index; S5, judging whether the process failure risk index exceeds a preset risk threshold; S6, when the process failure risk index exceeds the preset risk threshold, generating a new workflow topology structure and node task according to the latest case legal semantic feature vector and a preset legal compliance rule set; S7, continuously monitoring the execution of the new workflow topology structure and node task to calculate a process adaptability index; S8, judging whether the process adaptability index is lower than a preset adaptability threshold; S9, when the process adaptability index is lower than the preset adaptability threshold, generating a new workflow topology structure and node task again to form a closed-loop correction; S10, when the process adaptability index is not lower than the preset adaptability threshold, ending the process; S7 specifically comprises: The process execution effect quantitative value is evaluated by monitoring the task completion rate, the intermediate result of the legal risk report, and the matching degree with the latest case semantics, and the process adaptability index is calculated in combination with the process execution effect quantitative value and the risk index generated in the process of reconstructing the new workflow topology structure and node task; The process adaptability index combines the process execution effect quantitative value and the risk index possibly generated in the new workflow topology and node task reconstruction process; the calculation follows the following formula: ; wherein: : the flow adaptability index, whose technical meaning is the comprehensive trade-off between the new flow benefits and costs, is a scalar value normalized to the interval [0, 1], whose source is the calculation result of the present formula; : the flow execution effect quantitative value, its technical meaning is that the new flow in actual operation is consistent with the expected target, which is a normalized scalar value in the interval [0, 1]; its source is to monitor the execution of each node of the flow, the feedback of the mediator, and whether there is a new legal risk warning data, and the result is calculated by a preset weighted function; : the risk index that may be generated in the new process reconstruction process, its technical meaning is the estimation of potential problems brought by the complexity and uncertainty of the new process itself, which is a normalized scalar value in the interval [0, 1]; its source is that after the generation of the new process and before the formal execution, another AI evaluation model simulates and deduces it to evaluate the possible task allocation conflicts and legal clause omission risks; the AI evaluation model itself is obtained by supervised learning on the process structure features and actual failure mode data of historical reconstruction failure cases. 2.The AI large model-based mediation case intelligent analysis method according to claim 1, characterized in that, S2 specifically comprises: The multi-modal fusion large model extracts features of each modality data in the text materials, the speech transcription text, and the picture evidence of the case and embeds them to generate high-dimensional vector representations. 3.The AI large model-based mediation case intelligent analysis method of claim 1, wherein, S4 specifically comprises: The cosine similarity between the latest case legal semantic feature vector containing new evidence and the case core legal semantic feature vector is calculated, and the process failure risk index is calculated in combination with the quantitative value of the influence of the new evidence on the process. 4.The AI large model-based mediation case intelligent analysis method of claim 1, wherein, S6 specifically comprises: The process reconstruction model constructed based on the multi-modal fusion large model automatically generates or recommends a new workflow topology structure and node task according to the latest case legal semantic feature vector and the preset legal compliance rule set. 5.The AI large model-based mediation case intelligent analysis method of claim 1, wherein, S9 specifically comprises: The sensing-reconstruction cycle is triggered again to fine-tune or reconstruct the process until the process adaptability index reaches the preset adaptability threshold.

6. The mediation case intelligent analysis system based on an AI large model, applied to the mediation case intelligent analysis method based on an AI large model according to any one of claims 1 to 5, characterized in that, It comprises: A data input module for inputting the text materials, the speech transcription text, and the picture evidence of the case into the multi-modal fusion large model; A feature extraction module for extracting features of the text materials, the speech transcription text, and the picture evidence through the multi-modal fusion large model to generate high-dimensional vector representations; A semantic condensing module for deeply fusing and analyzing the high-dimensional vector representations to condense a case core legal semantic feature vector; A risk assessment module is configured to monitor a case core legal semantic feature vector, combine a latest case legal semantic feature vector containing new evidence analyzed by the multi-modal fusion large model, and calculate a process failure risk index. A process reconstruction module is configured to, when the process failure risk index exceeds a preset risk threshold, generate a brand-new workflow topology structure and node task according to the latest case legal semantic feature vector and a preset legal compliance rule set. An effect evaluation module is configured to continuously monitor the execution of the brand-new workflow topology structure and node task, and calculate a process adaptability index. A closed-loop correction module is configured to, when the process adaptability index is lower than a preset adaptability threshold, generate a brand-new workflow topology structure and node task again to form a closed-loop correction. 7.The AI large model-based mediation case intelligent analysis system according to claim 6, characterized in that, The application comprises a processor, a memory and a communication interface, the memory stores program code, and the processor is configured to execute the program code. ​

Images