Mediation case management method and device

Abstract

The invention provides a mediation case management method and device, and relates to the technical field of mediation, and the method comprises the steps: obtaining historical case information, to-be-mediated case information and mediator information; constructing a classification-matching model and a mediation scheme generation model; inputting the to-be-conciliated case information and the conciliator information into a classification-matching model to obtain a case type and a matching result of the to-be-conciliated case; inputting to-be-conciliated case information into a conciliation scheme generation model to obtain a preliminary conciliation scheme Constructing an AI digital human communication strategy according to the matching result and the preliminary mediation scheme; real-time video information is generated according to an AI digital person communication strategy, and a communication result is formed by combining real-time feedback information; and performing optimization processing on the initial mediation scheme according to a communication result to obtain a final mediation scheme and a final mediation case management report. The problems that the mediation efficiency is low during mediation case management, the information utilization rate is low, and attitude change of parties is not considered in generation of an A I communication strategy video are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of mediation. Specifically, it relates to a mediation case management method and device. Background Art

[0002] With the complexity of the economy and interpersonal relationships, various disputes occur frequently. As a method of resolving disputes, mediation is widely used in many fields such as family, business, and labor related to the economy. Mediation case management refers to the whole process of effectively organizing, allocating, tracking, handling, and feedback of various cases during the mediation process, involving multiple links, from the initial acceptance to the result tracking and optimization after the mediation ends, in order to achieve a fair resolution of disputes.

[0003] However, many mediation works still rely on manual entry, allocation, and tracking of case progress, which is not only time-consuming but also prone to errors. Secondly, there is a lack of unified standards and norms, and the professional capabilities and time of mediators when handling different cases are not taken into account, and it is overly dependent on the experience of mediators. When the case allocation is not ideal, it is difficult to achieve effective collaborative work and a high acceptance rate of the mediation plan by the parties. At the same time, there is currently a lack of analysis and feedback optimization of the handling of mediation cases, making it difficult for mediators to extract experience from historical data to guide future work. The current mediation case management has a lot of manual intervention and low efficiency, and it is also difficult to handle a large number of cases. Generating videos through AI communication strategies is an effective method to improve mediation efficiency, but the videos generated by existing AI communication strategies cannot cope with the drastic changes that may occur in the needs, emotions, and reactions of the parties in complex mediation scenarios, lacking the ability of rapid response and empathy, and unable to interact highly personalized like human mediators.

[0004] Therefore, there is an urgent need for an efficient, intelligent, and systematic management method and device to solve the problems of low mediation efficiency, low information utilization rate in mediation case management, and the generation of AI communication strategy videos not considering the attitude changes of the parties. Summary of the Invention

[0005] The purpose of the present invention is to provide a mediation case management method and device to improve the above problems. To achieve the above purpose, the technical solutions adopted by the present invention are as follows:

[0006] In a first aspect, the present application provides a mediation case management method, including:

[0007] An acquisition module, configured to acquire historical case information, information of cases to be mediated, and mediator information, where the historical case information includes the situation of closed cases, the mediation process, and the mediation result, the information of cases to be mediated includes the basic information and real-time status of the cases, and the mediator information includes the historical case handling situation and the current working status of the mediator;

[0008] A first construction module, configured to construct a classification-matching model and a mediation plan generation model according to the historical case information;

[0009] An assignment and matching module, configured to input the case information to be mediated and the mediator information into the classification-matching model for classification and mediator matching, so as to obtain the case type and matching result of the case to be mediated;

[0010] A plan generation module, configured to input the case type of the case to be mediated and the case information to be mediated into the mediation plan generation model to obtain a preliminary mediation plan;

[0011] A second construction module, configured to construct an AI digital human communication strategy according to the matching result and the preliminary mediation plan;

[0012] A real-time communication module, configured to generate real-time video information according to the AI digital human communication strategy, and combine the real-time feedback information of the case parties to form a communication result;

[0013] An optimization module, configured to perform optimization processing on the preliminary mediation plan according to the communication result to obtain a final mediation plan and a mediation case management report.

[0014] In a second aspect, the present application further provides a mediation case management device, including:

[0015] An acquisition module, configured to acquire historical case information, case information to be mediated, and mediator information, where the historical case information includes the situation of closed cases, the mediation process, and the mediation result, the case information to be mediated includes the basic information and real-time status of the case, and the mediator information includes the historical case handling situation and the current working status of the mediator;

[0016] A first construction module, configured to construct a classification-matching model and a mediation plan generation model according to the historical case information;

[0017] An assignment and matching module, configured to input the case information to be mediated and the mediator information into the classification-matching model for classification and mediator matching, so as to obtain the case type and matching result of the case to be mediated;

[0018] A plan generation module, configured to input the case type of the case to be mediated and the case information to be mediated into the mediation plan generation model to obtain a preliminary mediation plan;

[0019] A second construction module, configured to construct an AI digital human communication strategy according to the matching result and the preliminary mediation plan;

[0020] A real-time communication module, which is used to generate real-time video information according to the AI digital human communication strategy and form a communication result by combining the real-time feedback information of the case parties;

[0021] An optimization module, which is used to optimize the preliminary mediation plan according to the communication result to obtain a final mediation plan and a mediation case management report.

[0022] The beneficial effects of the present invention are as follows: By using a classification-matching model to combine case theme modeling and mediator historical data, the present invention ensures that cases are assigned to the most suitable mediators, improving the matching accuracy. And based on case characteristics and historical data, a mediation plan is intelligently generated, avoiding relying on manual decision-making and ensuring that the plan conforms to the repayment ability of the parties.

[0023] During the mediation process, through the combination of AI digital humans and multi-modal sentiment analysis, the communication strategy is dynamically adjusted to improve communication efficiency. At the same time, an optimization algorithm is used to optimize the mediation plan based on the sentiment feedback. Finally, through an incremental training mechanism, the mediation plan generation model is continuously optimized, improving the accuracy and intelligence level of future case processing. Therefore, the present invention greatly improves the mediation efficiency, reduces the burden on mediators, and increases the mediation success rate, providing an innovative solution for intelligent mediation.

[0024] Other features and advantages of the present invention will be described in the subsequent specification, and, in part, will become apparent from the specification or can be understood by implementing the embodiments of the present invention. The objectives and other advantages of the present invention can be achieved and obtained by the structures specifically pointed out in the written specification, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the embodiments. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0026] Figure 1 It is a schematic flow chart of the mediation case management method described in the embodiments of the present invention;

[0027] Figure 2 It is a schematic structural diagram of the mediation case management device described in the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Components of the embodiments of the present invention generally described and illustrated in the drawings herein can be arranged and designed in a variety of different configurations. Therefore, the detailed description of the embodiments of the present invention provided herein is not intended to limit the scope of the claimed invention, but is merely representative of selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0029] It should be noted that like reference numerals and letters denote like items in the following drawings. Therefore, once an item is defined in one drawing, it does not require further definition and explanation in subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive distinction and cannot be construed as indicating or implying relative importance.

[0030] Embodiment 1:

[0031] This embodiment provides a mediation case management method.

[0032] See Figure 1 , which shows that this method includes steps S1, S2, S3, S4, S5, S6 and S7.

[0033] Step S1: Obtain historical case information, information of the case to be mediated, and mediator information. The historical case information includes the situation of the closed cases, the mediation process and the mediation results. The information of the case to be mediated includes the basic information and the real-time status of the case. The mediator information includes the historical case handling situation and the current working status of the mediator;

[0034] In this embodiment, the situation of the closed cases includes the basic information of the closed cases, the designated mediator and the corresponding mediation case management report. Among them, the basic information is provided by both parties, including the information of both parties, financial information and case situation description information. The case situation description information includes the case background, case facts and case claims. The case background describes the cause of the dispute, as well as the time and place where the dispute occurred. The case facts describe the actions and transaction records of the parties, and provide evidence such as contract terms, performance situations and agreed matters. At the same time, a case number is generated for each case to facilitate continuous management and query.

[0035] Step S2: Construct a classification-matching model and a mediation plan generation model according to the historical case information;

[0036] Step S3: Input the information of the case to be mediated and the information of the mediator into the classification - matching model for classification and mediator matching, to obtain the case type of the case to be mediated and the matching result;

[0037] In this embodiment, since manual classification of case type information often relies on the experience and judgment of mediators, different mediators may make different classifications for the same case, which is likely to introduce biases and inconsistencies. And manual classification and mediation are often restricted by the personal professional knowledge and experience of mediators. If a mediator does not have sufficient domain knowledge, they may not be able to accurately judge the nature of the case and match a suitable mediator. At the same time, when assigning mediators, factors such as the professional expertise and workload of mediators are not considered, and cases cannot be reasonably allocated. Therefore, classifying the case type through the classification - matching model and assigning a suitable mediator improves the processing efficiency and breaks the knowledge limitation.

[0038] The said step S3 includes:

[0039] Step S31: Perform text pre - processing on the basic information, and obtain the pre - processed text by removing stop words, word segmentation, and part - of - speech tagging;

[0040] Step S32: Perform topic modeling on the pre - processed text to obtain the topic distribution of the case to be mediated;

[0041] In this embodiment, the content of the basic information is extensive. When there is information overload, it is difficult to manually screen out all key elements to obtain an accurate classification. In this step, topic modeling is used to automatically identify potential topics, discover words related to specific topics from the messy text, and reveal the core content of the text through these words. Therefore, topic modeling can identify the key areas and core issues of the case, providing a more accurate basis for case classification and mediator matching.

[0042] Step S33: Calculate the cosine similarity between the topic distribution of the case to be mediated and the pre - defined topic distribution of the case type, and select the pre - defined case type with the largest cosine similarity as the case type of the case to be mediated;

[0043] In this embodiment, a preset topic distribution is set for each case type, that is, it includes multiple topic keywords, and the number of topics in the topic distribution is set. Matching is performed by calculating the cosine similarity to obtain the case type of the case to be mediated.

[0044] Step S34: According to the current working status of the mediator, obtain the current number of mediated cases and the maximum load capacity of each mediator, and according to the historical case - handling situation of the mediator, obtain the historical domain information of each mediator;

[0045] In this embodiment, an initial priority index for each case type that a mediator is good at is set. Then, based on the historical case handling situation, the number of each case type handled by each mediator is counted and sorted in descending order. An incremental priority index corresponding to each case type is assigned according to the sorting position. This incremental priority index represents the mediator's handling experience for each case type. By adding the initial priority index and the incremental priority index, the priority index of each field that the mediator is good at is obtained.

[0046] Step S35: Screen mediators who meet the matching conditions for the case to be mediated to obtain multiple screened mediators. The matching conditions are that the historical field information of the mediator contains the case type of the case to be mediated and the current number of mediated cases of the mediator is less than the maximum load capacity.

[0047] Step S36: Calculate the priority score of each screened mediator for the case to be mediated based on the parameter information of the screened mediators.

[0048] In step S36, the calculation formula for the priority score is:

[0049]

[0050] In the formula, S represents the priority score, α1 and α2 represent weight parameters, P represents the priority index of the case type of the case to be mediated in the fields that the mediator is good at, C represents the current number of mediated cases of the mediator, and C max represents the maximum load capacity of the mediator.

[0051] In this embodiment, by considering the mediator's own situation and assigning the case types in the fields that the mediator is good at, the efficiency and quality of case handling can be improved, ensuring that the mediator can give full play to their professional advantages to the greatest extent. Also, the current workload of the mediator is considered, avoiding assigning too many cases to mediators with an overloaded workload, ensuring that the workload of each mediator is within a reasonable range, and preventing the decline in handling quality or mediator fatigue caused by overwork.

[0052] Step S37: Select the initially screened mediator with the highest priority score as the designated mediator for the case to be mediated to obtain the matching result.

[0053] Step S4: Input the case type of the case to be mediated and the information of the case to be mediated into the mediation plan generation model to obtain a preliminary mediation plan.

[0054] The said step S4 includes:

[0055] Step S41: Extract text information from the basic information to obtain text key information, where the text key information includes key case factors, expected repayment methods, repayment ability information, and historical repayment records;

[0056] In this embodiment, keywords and phrases are extracted from the basic information to obtain the total amount of this repayment, the expected repayment method of the parties, repayment ability information, and historical repayment records. Among them, the repayment ability information includes the income-to-debt ratio, asset-liability ratio, and monthly income, and the historical repayment records include whether the repayment is made on time, whether there is a default record, and the historical repayment amount. The repayment ability information is obtained from the basic information filled in by the parties, and the parties are reminded to fill in relevant information when establishing a mediation case, without involving privacy infringement.

[0057] Step S42: Calculate the priority factors of repayment types through the text key information, where the repayment types include lump-sum repayment, flexible repayment, and installment repayment;

[0058] The step S42 includes:

[0059] Step S421: Obtain the initial priority factor of each repayment type through the case type of the case to be mediated;

[0060] Step S422: Calculate the historical adjustment factor of each repayment type according to the historical repayment records;

[0061] In this embodiment, the number of on-time repayments and the number of defaults of each repayment type are counted through the historical repayment records, and the historical adjustment factor is calculated. Among them, the calculation formula of the historical adjustment factor is:

[0062]

[0063] In the formula, C h represents the historical adjustment factor, x t represents the total number of repayments, x h represents the number of on-time repayments, x d represents the number of defaults.

[0064] Step S423: Calculate the credit score through the repayment ability information and the historical repayment records, and calculate the default risk factor of each repayment type through the credit score;

[0065] In this embodiment, the default risk factor is an important indicator for evaluating whether the parties to the case may fail to fulfill the repayment obligation on time. The default risk factor is usually closely related to the credit score. The higher the credit score, the lower the default risk. Conversely, the lower the credit score, the higher the default risk.

[0066] The calculation formula of the credit score is:

[0067] C s = w1L f + w2I in + w3L cs + w4L h

[0068] Wherein, C s represents the credit score, w1, w2, w3, and w4 all represent weight coefficients, L f represents the debt ratio, I in represents the monthly income, L cs represents the credit report score, L h represents the number of defaults.

[0069] The calculation formula of the default risk factor is:

[0070]

[0071] Wherein, C w represents the default risk factor, C s represents the credit score, k represents the adjustment parameter to control the steepness of the curve, and θ represents the credit score threshold.

[0072] By fitting the default risk factors of each repayment type with historical data, the specific parameters in the formulas of the default risk factors of each repayment type are obtained. For different repayment types, the default risk factor of lump-sum repayment decreases rapidly with the increase of the credit score, which is suitable for parties with higher credit scores. While the default risk factors of flexible repayment and installment repayment decrease slowly with the increase of the credit score.

[0073] Step S424: Calculate the monthly repayment amount according to the repayment ability information and the preset repayment ability evaluation formula;

[0074] In this embodiment, the calculation formula of the monthly repayment amount is:

[0075] R m = min(R1, R2)

[0076] R1 = I in -(I d + I e )

[0077] Wherein, R m represents the monthly repayment amount, min(·) represents taking the minimum value, R1 represents the disposable income, I in represents the monthly income, I d represents the monthly debt, I e represents the basic living expenses, and R2 represents the credit limit.

[0078] Step S425: Calculate the ability matching factor for one-time repayment based on the monthly repayment amount and the total amount of this repayment;

[0079] In this embodiment, when the monthly repayment amount is greater than the total amount of this repayment, the ability matching factor for one-time repayment is 1; when the monthly repayment amount is less than the total amount of this repayment, the ability matching factor for one-time repayment is 0, indicating that one-time repayment cannot be completed.

[0080] Step S426: Obtain the priority factor for each repayment type based on the initial priority factor, the historical adjustment factor, the default risk factor, and the ability matching factor.

[0081] In this embodiment, the calculation formula for the priority factor is:

[0082] C T =(β1C a +β2C h +β3(1 - C w ))·C n

[0083] In the formula, C T represents the priority factor, β1, β2, and β3 represent adjustment parameters, C a represents the initial priority factor, C h represents the historical adjustment factor, C w represents the default risk factor, C n represents the ability matching factor, and the ability matching factors for flexible repayment and installment repayment are 1.

[0084] Step S43: Perform one-hot encoding and standardization processing on the text key information and the priority factors of the repayment types to obtain the encoded feature information;

[0085] In this embodiment, one-hot encoding is performed on the categorical data, creating a new binary column for each category to obtain a binary feature. At the same time, standardization processing is performed on the numerical data, converting it into a distribution with a mean of 0 and a standard deviation of 1 to eliminate the dimensional differences between different features.

[0086] Step S44: Input the encoded feature information into the trained XGBoost model to obtain the first mediation plan;

[0087] In this embodiment, a mediation plan is obtained through the XGBoost model. It can handle problems such as missing values and imbalanced data, and is less sensitive to data distribution and feature selection. Since the information of each case is provided by the corresponding parties, there may be incomplete data. However, the XGBoost model has stronger robustness when dealing with various complex data in mediation cases. At the same time, it can handle complex non-linear relationships, so it can capture the complex relationship between the key factors of the case and the repayment type.

[0088] Step S45: A preliminary mediation plan is obtained by a designated mediator to confirm and adjust the first mediation plan.

[0089] In this embodiment, by intelligently generating a mediation plan and also through a designated mediator for manual intervention, using their own experience and professional level, the first mediation plan generated by intelligence is confirmed and adjusted to obtain a reasonable preliminary mediation plan.

[0090] Step S5: Construct an AI digital human communication strategy based on the matching result and the preliminary mediation plan;

[0091] In this embodiment, a personalized AI digital human communication strategy is generated for each case. According to the matched mediator information, the personality characteristics, communication style, and language expression of the mediator can be simulated. For example, if the mediator's style is rational and concise, the AI digital human communication strategy will also tend to be straightforward. When the mediator is more inclined to empathy and delicate expression, the AI digital human communication strategy will simulate a more caring and encouraging tone.

[0092] The communication method of the AI digital human communication strategy is also adjusted according to the mediator's historical records, such as mediation success rate, case types handled, etc., to conform to the best performance method of the mediator.

[0093] At the same time, the AI digital human communication strategy needs to be able to provide legal opinions or bases related to the case. Therefore, necessary legal provisions, interpretations, and references to relevant cases can be obtained from a preset legal library to ensure the legality and compliance of the mediation process.

[0094] According to the preliminary mediation plan of the case, the AI digital human communication strategy can also generate mediation suggestions that conform to the current situation. For example, if the repayment method is installment, the AI digital human communication strategy can simulate the mediator to discuss a reasonable installment repayment plan with the parties and provide auxiliary information such as repayment ability assessment at the same time.

[0095] Step S6: Generate real-time video information according to the AI digital human communication strategy and form a communication result in combination with the real-time feedback information of the case parties;

[0096] The said step S6 includes:

[0097] Step S61: Generate real-time video information according to the AI digital human communication strategy;

[0098] Step S62: Obtain the real-time feedback information of the case parties on the real-time video information;

[0099] In this embodiment, feedback information is obtained in real time through multi-modal data such as video and voice, breaking through the limitation of a single feedback form on subsequent sentiment analysis.

[0100] Step S63: Perform multi-modal sentiment analysis on the real-time feedback information to obtain the state transition information of the case parties, where the state transition information includes emotion type transition information and emotion intensity transition information;

[0101] The step S63 includes:

[0102] Step S631: Obtain video feedback information, audio feedback information, and text feedback information through the real-time feedback information;

[0103] Step S632: Extract facial features from the video feedback information through a facial detection algorithm;

[0104] Step S633: Perform facial expression recognition on the facial features through an expression recognition algorithm and classify them according to preset emotion labels to obtain facial expression information;

[0105] Step S634: Extract audio features from the audio feedback information to obtain acoustic features, where the acoustic features include pitch features, speech rate features, and speech energy features;

[0106] Step S635: Perform emotion classification on the acoustic features through a speech emotion recognition algorithm to obtain speech emotion information;

[0107] Step S636: Perform emotion vocabulary recognition on the text feedback information based on natural language processing methods and emotion dictionaries to obtain emotion vocabulary;

[0108] Step S637: Perform emotion classification on the emotion vocabulary through an emotion vocabulary library to obtain text emotion information;

[0109] Step S638: Perform multi-modal feature fusion on the facial expression information, speech emotion information, and text emotion information to obtain multi-modal emotion features;

[0110] Step S639: Perform emotion state transition detection on the multi-modal emotion features through an LSTM neural network, capture emotion state changes, and obtain the state transition information of the case parties.

[0111] In this embodiment, during the traditional mediation process, the analysis of emotional states mostly relies on manual judgment, which is prone to overlooking and being affected by subjective factors. Through multimodal emotion analysis, emotional changes can be objectively identified and tracked, thereby providing more accurate feedback and adjustment. Moreover, by combining multimodal data of video, audio, and text for comprehensive emotion analysis, the AI digital human can understand and adapt to the emotional needs of the parties more accurately than single-modal analysis.

[0112] Therefore, it is possible to deeply explore the emotional changes of the parties in the video and achieve all-round emotion analysis by combining audio and text information. Through the detection of changes in emotional states, the AI digital human can timely adjust its expression mode and optimize the mediation process.

[0113] Step S64: Extract the scheme adjustment opinions in the text feedback information of the real-time feedback information through natural language processing methods and the trained intention recognition model;

[0114] In this embodiment, key information is extracted through natural language processing methods, and then the intention recognition model is used to determine the needs of the customer, such as extending the repayment period and adjusting the repayment amount, etc. Among them, the intention recognition model can use the BERT model or the RoBERTa model, which can process long text context information.

[0115] Step S65: Obtain the communication result based on the scheme adjustment opinion and the state transition information.

[0116] Step S7: Optimize the preliminary mediation scheme according to the communication result to obtain the final mediation scheme and the mediation case management report.

[0117] The said Step S7 includes:

[0118] Step S71: Define a particle swarm and initialize the particle swarm with the preliminary mediation scheme. Each particle in the particle swarm represents a mediation scheme;

[0119] Step S72: Obtain the emotion score according to the communication result and update the constraint conditions according to the communication result. The constraint conditions include repayment time constraint, repayment amount constraint, and repayment ability constraint;

[0120] Step S73: Construct an objective function through the emotion score, the repayment ability matching score of the case parties, and the default risk score of the case parties;

[0121] Step S74: Calculate the fitness of each particle through the objective function and update all particles according to the fitness until the iteration times are reached, and select the particle with the maximum fitness as the optimized mediation scheme;

[0122] Step S75: Adjust the AI digital human communication strategy based on the optimized mediation plan and the status transition information, and update the real-time video information until the parties to the case stop communicating;

[0123] Step S76: Conduct manual mediation through the final optimized mediation plan to obtain the final mediation plan and the mediation case management report.

[0124] In this embodiment, through the real-time communication between the parties and the AI digital human, the mediation plan is updated in real time, and the repayment type, repayment time, and amount of each repayment are adjusted, so as to improve the satisfaction of the parties.

[0125] The said Step S76 includes:

[0126] Step S761: Obtain the mediation process of the AI digital human and the mediation result of the AI digital human for the case to be mediated. The mediation result of the AI digital human includes whether the mediation is successful and the final optimized mediation plan;

[0127] Step S762: Based on the mediation process of the AI digital human, the mediation result of the AI digital human, and the matching result, conduct manual mediation to obtain the final mediation result. The final mediation result includes whether the mediation is successful and the final mediation plan;

[0128] Step S763: Generate a mediation case management report through the final mediation result and all the mediation processes. The mediation case management report is used to perform incremental training on the mediation plan generation model to obtain an optimized mediation plan generation model.

[0129] In this embodiment, when the mediation of the AI digital human is completed, the mediation result may be successful or failed. If it is successful, the case is directly closed and a mediation case management report is generated. If the mediation of the AI digital human fails, manual mediation is conducted by the designated mediator who is matched. If the manual mediation also fails, the parties may be advised to file a lawsuit or make an appointment for the next manual mediation. The designated mediator also makes a manual record of the mediation process during the manual mediation and finally generates a mediation case management report.

[0130] Meanwhile, the successfully mediated cases among the closed cases are used as new samples for incremental training to fine-tune the mediation plan generation model. Since a large number of new cases are generated every day, through incremental training, it is not necessary to retrain the entire model, which can quickly adapt to the latest data, significantly reduce the occupation of computing resources, and improve the training efficiency.

[0131] To sum up, the present invention designs a classification-matching model, calculates the matching degree based on the case theme modeling and the historical data of the mediators, and ensures that the case is assigned to the most suitable mediator. It can avoid inexperienced mediators handling complex cases and improve the success rate.

[0132] Key information of the case is also extracted, and a model is trained based on historical case data to automatically generate an optimal preliminary mediation plan, avoiding the need to rely entirely on manual plan formulation and improving mediation efficiency. At the same time, it can accurately match the capabilities of the parties. By combining historical data and credit assessment, reasonable repayment methods are screened to ensure that the plan meets the needs of the parties without exceeding the tolerance of the other party.

[0133] An AI digital human communication strategy is designed to simulate the style of mediators, and different tones and expressions are used according to the case situation to improve communication effects. The plan is dynamically adjusted in combination with the feedback of the parties to make the plan more personalized and acceptable. At the same time, multi-modal sentiment analysis is carried out, integrating facial expressions, voice features and text analysis to detect the emotional fluctuations of the parties in real time. The repayment plan is optimized based on the sentiment analysis to ensure the feasibility and optimal solution of the mediation plan. The XGBoost model is continuously optimized using the data of closed cases to improve the mediation effect of future cases.

[0134] Therefore, the present invention effectively improves the efficiency, transparency and standardization of mediation work, and provides strong technical support for the modernization of the mediation industry. It is not only applicable to various types of mediation cases, but also provides the possibility of continuous optimization for future mediation work, and has broad application prospects.

[0135] Embodiment 2:

[0136] As Figure 2 shown, this embodiment provides a mediation case management device, and the device includes:

[0137] An acquisition module, configured to acquire historical case information, information of the case to be mediated, and mediator information, where the historical case information includes the situation of closed cases, the mediation process and the mediation result, the information of the case to be mediated includes the basic information and the real-time status of the case, and the mediator information includes the historical case handling situation and the current working status of the mediator;

[0138] A first construction module, configured to construct a classification-matching model and a mediation plan generation model according to the historical case information;

[0139] An assignment and matching module, configured to input the information of the case to be mediated and the mediator information into the classification-matching model for classification and mediator matching to obtain the case type and matching result of the case to be mediated;

[0140] A plan generation module, configured to input the case type of the case to be mediated and the information of the case to be mediated into the mediation plan generation model to obtain a preliminary mediation plan;

[0141] A second construction module, configured to construct an AI digital human communication strategy according to the matching result and the preliminary mediation plan;

[0142] A real-time communication module, which is used to generate real-time video information according to the AI digital human communication strategy, and combine the real-time feedback information of the case parties to form a communication result;

[0143] An optimization module, which is used to optimize the preliminary mediation plan according to the communication result to obtain the final mediation plan and the mediation case management report.

[0144] The allocation and matching module includes:

[0145] A preprocessing unit, which preprocesses the text of the basic information to obtain preprocessed text by removing stop words, word segmentation, and part-of-speech tagging;

[0146] A theme extraction unit, which performs theme modeling on the preprocessed text to obtain the theme distribution of the case to be mediated;

[0147] A first calculation unit, which is used to calculate the cosine similarity between the theme distribution of the case to be mediated and the pre-defined case type theme distribution, and select the pre-defined case type with the largest cosine similarity as the case type of the case to be mediated;

[0148] An acquisition unit, which is used to obtain the current number of mediation cases and the maximum load capacity of each mediator according to the current working status of the mediator, and obtain the historical field information of each mediator according to the historical case handling situation of the mediator;

[0149] A screening unit, which is used to screen mediators who meet the matching conditions for the case to be mediated to obtain multiple screened mediators, and the matching conditions are that the historical field information of the mediator contains the case type of the case to be mediated and the current number of mediation cases of the mediator is less than the maximum load capacity;

[0150] A second calculation unit, which is used to calculate the priority score of each screened mediator for the case to be mediated based on the parameter information of the screened mediator;

[0151] A matching unit, which is used to select the initially screened mediator with the highest priority score as the designated mediator for the case to be mediated to obtain a matching result.

[0152] The plan generation module includes:

[0153] An information extraction unit, which is used to extract text information from the basic information to obtain text key information, and the text key information includes case key factors, expected repayment methods, repayment ability information, and historical repayment situations;

[0154] A third calculation unit, configured to calculate a priority factor of a repayment type through the key text information, where the repayment types include lump-sum repayment, flexible repayment, and installment repayment;

[0155] An encoding processing unit, configured to perform one-hot encoding and normalization processing on the key text information and the priority factor of the repayment type to obtain encoded feature information;

[0156] A solution acquisition unit, configured to input the encoded feature information into a trained XGBoost model to obtain a first mediation solution;

[0157] A confirmation unit, configured to confirm and adjust the first mediation solution by a designated mediator to obtain a preliminary mediation solution.

[0158] It should be noted that regarding the device in the above embodiments, the specific manners in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated herein.

[0159] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

[0160] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present invention, and all should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A mediation case management method, characterized in that, Including: Obtain historical case information, information of the case to be mediated, and mediator information. The historical case information includes the situation of the closed case, the mediation process, and the mediation result. The information of the case to be mediated includes the basic information and the real-time status of the case. The mediator information includes the historical case handling situation and the current working status of the mediator; Construct a classification-matching model and a mediation plan generation model based on the historical case information; Input the information of the case to be mediated and the mediator information into the classification-matching model for classification and mediator matching, and obtain the case type and matching result of the case to be mediated; Input the case type of the case to be mediated and the information of the case to be mediated into the mediation plan generation model to obtain a preliminary mediation plan; Construct an AI digital human communication strategy based on the matching result and the preliminary mediation plan; Generate real-time video information according to the AI digital human communication strategy, and combine the real-time feedback information of the case parties to form a communication result; Optimize the preliminary mediation plan according to the communication result to obtain the final mediation plan and the mediation case management report.

2. The mediation case management method according to claim 1, wherein , The inputting the information of the case to be mediated and the mediator information into the classification-matching model for classification and mediator matching, and obtaining the matching result includes: Perform text preprocessing on the basic information, and obtain the preprocessed text by removing stop words, word segmentation, and part-of-speech tagging; Perform topic modeling on the preprocessed text to obtain the topic distribution of the case to be mediated; Calculate the cosine similarity between the topic distribution of the case to be mediated and the predefined case type topic distribution, and select the predefined case type with the largest cosine similarity as the case type of the case to be mediated; According to the current working status of the mediator, obtain the current number of mediation cases and the maximum load capacity of each mediator, and obtain the historical field information of each mediator according to the historical case handling situation of the mediator; Screen mediators who meet the matching conditions for the case to be mediated to obtain multiple screened mediators. The matching conditions are that the historical field information of the mediator contains the case type of the case to be mediated and the current number of mediation cases of the mediator is less than the maximum load capacity; Calculate the priority score of each screened mediator for the case to be mediated based on the parameter information of the screened mediators; Select the preliminary screened mediator with the highest priority score as the designated mediator of the case to be mediated to obtain the matching result.

3. The mediation case management method according to claim 1, wherein , The inputting the case type of the case to be mediated and the information of the case to be mediated into the mediation plan generation model to obtain a preliminary mediation plan includes: Extract text information from the basic information to obtain text key information. The text key information includes case key factors, expected repayment methods, repayment ability information, and historical repayment situations; Calculate the priority factor of the repayment type through the text key information. The repayment types include lump-sum repayment, flexible repayment, and installment repayment; Perform one-hot encoding and standardization processing on the text key information and the priority factor of the repayment type to obtain encoded feature information; Input the encoded feature information into the trained XGBoost model to obtain the first mediation plan; Have the designated mediator confirm and adjust the first mediation plan to obtain the preliminary mediation plan.

4. The mediation case management method according to claim 3, wherein , The calculating of the priority factor of the repayment type through the key text information includes: Obtain the initial priority factor of each repayment type through the case type of the case to be mediated; Calculate the historical adjustment factor of each repayment type according to the historical repayment situation; Calculate the credit score through the repayment ability information and the historical repayment situation, and calculate the default risk factor of each repayment type through the credit score; Calculate the monthly repayment amount according to the repayment ability information and the preset repayment ability evaluation formula; Calculate the ability matching factor for one-time repayment through the monthly repayment amount and the total amount of this repayment; Obtain the priority factor of each repayment type through the initial priority factor, the historical adjustment factor, the default risk factor and the ability matching factor.

5. The mediation case management method according to claim 1, wherein , The generating of the real-time video information according to the AI digital human communication strategy and combining with the real-time feedback information of the case parties to form the communication result includes: Generate real-time video information according to the AI digital human communication strategy; Obtain the real-time feedback information of the case parties on the real-time video information; Perform multi-modal sentiment analysis on the real-time feedback information to obtain the state change information of the case parties, and the state change information includes the emotion type change information and the emotion intensity change information; Extract the plan adjustment opinions in the text feedback information of the real-time feedback information through the natural language processing method and the trained intention recognition model; Obtain the communication result through the plan adjustment opinions and the state change information.

6. The mediation case management method according to claim 5, characterized in that , The performing of multi-modal sentiment analysis on the real-time feedback information to obtain the state change information of the case parties includes: Obtain video feedback information, audio feedback information and text feedback information through the real-time feedback information; Extract the facial features in the video feedback information through the face detection algorithm; Perform facial expression recognition on the facial features through the expression recognition algorithm and classify them according to the preset emotion labels to obtain the facial expression information; Extract the audio features of the audio feedback information to obtain the acoustic features, and the acoustic features include pitch features, speech rate features and speech energy features; Perform emotion classification on the acoustic features through the speech emotion recognition algorithm to obtain the speech emotion information; Perform emotion vocabulary recognition on the text feedback information based on the natural language processing method and the emotion dictionary to obtain the emotion vocabulary; Perform emotion classification on the emotion vocabulary through the emotion vocabulary library to obtain the text emotion information; Perform multi-modal feature fusion on the facial expression information, speech emotion information and text emotion information to obtain the multi-modal emotion features; Perform emotion state change detection on the multi-modal emotion features through the LSTM neural network to capture the emotion state change and obtain the state change information of the case parties.

7. The mediation case management method according to claim 5, characterized in that , The optimizing the preliminary mediation plan according to the communication result to obtain the final mediation plan and the mediation case management report includes: Define a particle swarm and initialize the particle swarm through a preliminary mediation plan. Each particle in the particle swarm represents a mediation plan; Obtain the emotional score based on the communication result, and update the constraint conditions according to the communication result. The constraint conditions include repayment time constraint, repayment amount constraint, and repayment ability constraint; Construct an objective function based on the emotional score, the repayment ability matching score of the case parties, and the default risk score of the case parties; Calculate the fitness of each particle through the objective function, and update all particles according to the fitness until the iteration times are reached. Select the particle with the maximum fitness as the optimized mediation plan; Adjust the AI digital human communication strategy based on the optimized mediation plan and the state transition information, and update the real-time video information until the case parties stop communicating; Conduct manual mediation through the final optimized mediation plan to obtain the final mediation plan and the mediation case management report.

8. A mediation case management device, characterized in that, Including: An acquisition module for acquiring historical case information, information of the case to be mediated, and mediator information. The historical case information includes the situation of the closed cases, the mediation process, and the mediation result. The information of the case to be mediated includes the basic information and the real-time status of the case. The mediator information includes the historical case handling situation and the current working status of the mediator; A first construction module for constructing a classification-matching model and a mediation plan generation model according to the historical case information; An allocation and matching module for inputting the information of the case to be mediated and the mediator information into the classification-matching model for classification and mediator matching to obtain the case type and the matching result of the case to be mediated; A plan generation module for inputting the case type of the case to be mediated and the information of the case to be mediated into the mediation plan generation model to obtain a preliminary mediation plan; A second construction module for constructing an AI digital human communication strategy according to the matching result and the preliminary mediation plan; A real-time communication module for generating real-time video information according to the AI digital human communication strategy and combining the real-time feedback information of the case parties to form a communication result; An optimization module for optimizing the preliminary mediation plan according to the communication result to obtain the final mediation plan and the mediation case management report.

9. The mediation case management method according to claim 8, characterized in that , the allocation and matching module includes: A preprocessing unit for preprocessing the text of the basic information, and obtaining the preprocessed text by removing stop words, word segmentation, and part-of-speech tagging; A theme extraction unit for performing theme modeling on the preprocessed text to obtain the theme distribution of the case to be mediated; A first calculation unit for calculating the cosine similarity between the theme distribution of the case to be mediated and the predefined case type theme distribution, and selecting the predefined case type with the maximum cosine similarity as the case type of the case to be mediated; An acquisition unit for obtaining the current number of cases being mediated and the maximum load capacity of each mediator according to the current working status of the mediator, and obtaining the historical field information of each mediator according to the historical case handling situation of the mediator; A screening unit, configured to screen mediators meeting the matching criteria for the case to be mediated, so as to obtain multiple screened mediators, where the matching criteria are that the historical field information of the mediator contains the case type of the case to be mediated and the current number of mediation cases of the mediator is less than the maximum load capacity; A second calculation unit, configured to calculate the priority score of each screened mediator for the case to be mediated based on the parameter information of the screened mediators; A matching unit, configured to select the initially screened mediator with the highest priority score as the designated mediator for the case to be mediated, so as to obtain a matching result.

10. The mediation case management method according to claim 8, characterized in that , The solution generation module includes: An information extraction unit, configured to perform text information extraction on the basic information to obtain text key information, where the text key information includes case key factors, expected repayment methods, repayment ability information, and historical repayment situations; A third calculation unit, configured to calculate the priority factor of the repayment type through the text key information, where the repayment types include lump-sum repayment, flexible repayment, and installment repayment; An encoding processing unit, configured to perform one-hot encoding and normalization processing on the text key information and the priority factor of the repayment type to obtain encoded feature information; A solution acquisition unit, configured to input the encoded feature information into a trained XGBoost model to obtain a first mediation solution; A confirmation unit, configured to confirm and adjust the first mediation solution through the designated mediator to obtain a preliminary mediation solution.

Images