A contradiction dispute mediation efficiency evaluation system

Abstract

The application discloses a kind of contradictory dispute mediation efficiency evaluation system, belong to the field of efficiency evaluation system, including setting weight and score: first in the system the weight score of factor and grade is set well;Mark: mark factor and grade in system case, mark result is fed into model and is trained;Deep learning model training: the result marked is trained in the model of fusion, i.e. data is trained in UIE big model;Prediction reasoning: after obtaining the prediction result by deep learning model, the final efficiency prediction score is obtained by self-defined algorithm operation, through the algorithm model based on UIE big model and CNN convolutional neural network.The model combines the advantages of user information extraction and convolutional neural network, can provide fast and accurate efficiency evaluation in the process of contradictory dispute mediation, make multi-dimensional judgment, evaluate case mediation efficiency score, deduce feedback from the processing flow of case, so as to evaluate the rationality of case handling.

Description

Technical Field

[0001] This invention relates to the field of performance evaluation systems, and in particular to a performance evaluation system for resolving conflicts and disputes. Background Technology

[0002] In today's society, with rapid economic development and profound changes in social structure, people's values, interests, and behaviors have undergone significant transformations. While these changes have promoted social development, they have also brought about a series of new social contradictions and disputes. The diversity, complexity, and uncertainty of these contradictions and disputes pose new challenges to traditional social management and conflict resolution mechanisms.

[0003] Against this backdrop, traditional conflict resolution mechanisms are proving inadequate. On the one hand, manual mediation is inefficient and struggles to cope with the ever-increasing number of conflicts; on the other hand, traditional methods often lack scientific rigor and systematic approach, failing to provide in-depth analysis of the root causes and development trends of conflicts. Furthermore, due to the complexity and diversity of conflicts and disputes, traditional solutions are often inefficient and ill-suited to the ever-changing social environment. Moreover, traditional methods fail to provide effective feedback after resolution, making it impossible to assess the reasonableness of each case's outcome. Therefore, establishing a more efficient, scientific, and fair conflict resolution mechanism has become a crucial issue for social governance innovation. Summary of the Invention

[0004] The main objective of this invention is to provide a system for evaluating the effectiveness of conflict resolution, which can effectively solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A system for evaluating the effectiveness of conflict resolution includes:

[0007] (a) Setting weights and scores: First, set the weights and scores of factors and levels in the system;

[0008] (ii) Labeling: Label the factors and levels in the system cases, and feed the labeling results into the model for training;

[0009] (III) Deep learning model training: The labeled results are trained in the fused model, that is, the data is trained in the UIE large model;

[0010] (iv) Predictive reasoning: After obtaining the prediction results through the deep learning model, the final performance prediction score is obtained through a custom algorithm.

[0011] Preferably, the factors and levels refer to the diversion effectiveness factors and diversion effectiveness levels, respectively, and both can be queried, modified, and exported within the system.

[0012] Preferably, the factors for improving the effectiveness of mediation include: degree of achievement of demands, emotional satisfaction, time to resolve conflicts, frequency of mediation, and scope of administrative intervention, with corresponding weights of 20%, 30%, 20%, 15%, and 15%, respectively.

[0013] Preferably, the diversion efficiency level is divided into: low, relatively low, average, relatively high and high, with corresponding score ranges of: 0~20, 21~40, 41~60, 61~80 and 81~100 respectively.

[0014] Preferably, the annotation result refers to the annotated content of the dataset, which is input in text form.

[0015] Preferably, the custom algorithm operation is based on the CNN convolutional neural network model, and the final output is a score and a text format result. The CNN convolutional neural network can extract the features of the data and realize the hierarchical representation and learning of the features. Secondly, different weights are set for the five dimensions of the mediation effectiveness: the degree of achievement of demands, emotional satisfaction, conflict resolution time, frequency of conflict mediation, and scope of administrative intervention.

[0016] Preferably, the custom algorithm is a custom algorithm that mainly integrates a CNN convolutional neural network model and a weighted multiplication algorithm; the specific steps are as follows:

[0017] ① Traverse the information of the five dimensions: The prediction data obtained through deep learning will contain information of different dimensions. It is necessary to first obtain the keywords of each dimension factor in the case.

[0018] ② CNN Convolutional Neural Network Classification: Keywords for each dimension factor are input into a CNN convolutional neural network for classification, allowing the machine to automatically identify the performance level of the keyword;

[0019] ③ Weighted multiplication algorithm: The effectiveness level and corresponding factor of the factors in different dimensions are automatically identified in the CNN convolutional neural network. The score is multiplied by the weight of the corresponding factor, and then the scores of different factors after being multiplied by the weight are added together to obtain the final total score. This score is the total effectiveness score of the case.

[0020] ④ Determine the level within the interval: The total performance score will be determined within a pre-set interval to determine the performance level of that score.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] This invention employs an algorithmic model based on the UIE large-scale model and CNN convolutional neural networks. This model combines the advantages of user information extraction and convolutional neural networks, aiming to quickly and accurately assess the effectiveness of conflict resolution. It provides rapid and accurate performance evaluation during the conflict resolution process, performs multi-dimensional judgments, evaluates the effectiveness score of case resolution, and provides feedback on the case handling process to assess the rationality of case handling. This system utilizes advanced artificial intelligence technology to comprehensively evaluate multi-dimensional factors of conflicts, enabling it to quickly and accurately predict and infer the effectiveness score of conflict resolution. This not only improves the efficiency and quality of conflict resolution but also helps to discover the underlying causes of conflicts, providing a scientific basis for formulating more effective resolution strategies. The development and application of this system will greatly improve the efficiency and quality of social governance, providing solid technical support for social harmony and stability. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating the effectiveness evaluation system for conflict resolution according to the present invention.

[0024] Figure 2 This is a schematic diagram of the factor analysis model for resolving conflicts and disputes according to the present invention;

[0025] Figure 3 This is a schematic diagram of the CNN convolutional neural network classification process of the present invention;

[0026] Figure 4 This is a schematic diagram of the weighted multiplication algorithm in the custom algorithm of this invention. Detailed Implementation

[0027] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0028] like Figure 1 As shown, a conflict resolution effectiveness evaluation system includes:

[0029] (a) Setting weights and scores: First, set the weights and scores of factors and levels in the system; factors and levels refer to the guidance effectiveness factors and guidance effectiveness levels, respectively, and both can be queried, modified and exported in the system.

[0030] The factors contributing to the effectiveness of mediation include: degree of grievance fulfillment, emotional satisfaction, time to resolve conflict, frequency of mediation, and scope of administrative intervention, with corresponding weights of 20%, 30%, 20%, 15%, and 15%, respectively. The effectiveness levels are categorized as low, relatively low, average, relatively high, and high, with corresponding score ranges of 0–20, 21–40, 41–60, 61–80, and 81–100, respectively.

[0031] (ii) Labeling: Label the factors and levels in the system case, and feed the labeling results into the model for training; the labeling results refer to the labeled content of the dataset, which is input in text form.

[0032] like Figure 2 As shown, based on primary research data and a database of conflict and dispute cases, a factor analysis model for conflict resolution was established using grounded theory selective coding to establish relevant indicators. This model incorporates opinions from experts and practitioners, and verifies structural integrity and process coverage through randomly selected case texts. The model aims to explain, from the perspective of the game between the parties involved in the conflict and the government, the impact of the game's motivations, resource constraints, and environmental regulations on the methods and effectiveness of conflict resolution, based on the results of selective coding. The parties involved and the relevant government departments, as the two main actors, occupy opposite ends of a game balance, engaging in a game based on their respective considerations and corresponding pressures, shaping their ability to resolve conflicts. The direction of this game ultimately affects whether the risks of conflict and dispute events can be effectively controlled, and whether it will shift from routine resolution to more complex and high-risk abnormal resolution. Environmental regulations form the foundation and action framework of the game between the two main actors. The strength of legal and policy regulations, organizational environmental regulations, and social environmental regulations significantly influences the effectiveness of conflict resolution. More robust environmental regulations can help control conflict risks. The resource constraints of game participants characterize their ability to take action. Government departments choosing routine mediation can save resources in action, information, and negotiation; conversely, unorthodox mediation incurs greater costs. This model, based on grounded theory coding, considers both indicator importance and data availability. It selects five main categories: game dynamics, game capabilities, conflict resolution methods, and environmental regulation. From these categories influencing conflict resolution effectiveness, it extracts representative and quantifiable indicators. Ultimately, it selects five factors: the degree to which the parties' demands are met, emotional satisfaction, and the time, frequency, and scope of administrative intervention by government departments. A five-equal-weight scale is used to classify and comprehensively evaluate the final effectiveness of conflict resolution work. The quantified effectiveness evaluation results, generated through subjective and objective scoring and weighted scoring, can be used to measure the resource costs incurred by relevant units in achieving mediation effectiveness and predict the risk of further escalation of the conflict, thereby guiding and optimizing conflict resolution methods.

[0033] (III) Deep learning model training: The labeled results are used to train the fused model. When training the fused model, attention should be paid to the training time, because the capacity of the fused model is large and the code is more complex. The training time of the data in the UIE large model is relatively long.

[0034] (iv) Predictive Inference: After obtaining the prediction results through the deep learning model, a custom algorithm is then used to calculate the final performance prediction score. The custom algorithm is based on the CNN convolutional neural network model and outputs the score and the result in text format.

[0035] like Figure 3 - Figure 4 As shown, the custom algorithm in this model is primarily a fusion of a CNN convolutional neural network model and a weighted multiplication algorithm. The CNN convolutional neural network can extract data features and achieve hierarchical representation and learning of these features. Secondly, different weights are assigned to the five dimensions of mediation effectiveness: achievement of demands, emotional satisfaction, conflict resolution time, frequency of conflict mediation, and scope of administrative intervention. Each of the five dimensions has five effectiveness levels: low, relatively low, average, relatively high, and high. These five dimensions and effectiveness levels form a matrix with 25 corresponding scores. The specific steps are as follows:

[0036] ① Traverse the information of the five dimensions: The prediction data obtained through deep learning will contain information of different dimensions. It is necessary to first obtain the keywords of each dimension factor in the case.

[0037] ② CNN Convolutional Neural Network Classification: Keywords for each dimension factor are input into a CNN convolutional neural network for classification, allowing the machine to automatically identify the performance level of the keyword;

[0038] ③ Weighted multiplication algorithm: The effectiveness level and corresponding factor of the factors in different dimensions are automatically identified in the CNN convolutional neural network. The score is multiplied by the weight of the corresponding factor, and then the scores of different factors after being multiplied by the weight are added together to obtain the final total score. This score is the total effectiveness score of the case.

[0039] ④ Determine the level within the interval: The total performance score will be determined within a pre-set interval to determine the performance level of that score.

[0040] The main functions used are word splitting and CNN (Convolutional Neural Network). Word splitting aims to transform the dataset into language that the computer can understand, making it easier to feed into the CNN for training. CNN is a deep learning model that can be used for text classification in NLP (Natural Language Processing). It converts text into word vectors and applies convolution and pooling operations to capture local features in the text, thereby modeling and classifying the text.

[0041] 1. Input layer: The input layer concatenates the word vectors of all words (padding) in a sentence into a matrix. The word vectors change as the model is trained.

[0042] 2. Convolutional Layers: Each convolutional kernel has a size of filter_size * embedding_size. filter_size represents the number of words contained in the kernel along its vertical axis, assuming a word order relationship between adjacent words. embedding_size is the dimension of the word vectors. After each convolutional kernel completes its computation, we obtain a column vector representing the features extracted from the sentence by that kernel. The number of convolutional kernels determines the number of feature types extracted.

[0043] 3. Pooling layer: In the pooling layer, the pooling operation extracts the maximum value of the column vectors obtained from the convolution and compresses it to reduce the number of features.

[0044] 4. Fully Connected Layer: To transform the vector output by the pooling layer into the desired prediction result, a softmax layer is added. For example, in a classification task involving factor performance levels, this involves transforming the vector into different level results.

[0045] It should be noted that the score settings are configured in the rating score settings interface. Input data and label the dataset. During training, this labeled content can be fed into the model for training. The final result is obtained through the algorithm model. The model prediction time is approximately 4000 milliseconds. Multiple tests have verified that the predicted rating in the performance evaluation is consistent with the actual case description, with no discrepancies. It is important to note that the model runs relatively stably under Python 3.8.1, NumPy version 1.24.3, and PaddlePaddle version 2.5.2. In the deployment environment, the choice of NumPy and PaddlePaddle versions is crucial; incompatible versions can easily lead to system errors. Additionally, PaddleNLP version 2.5.2 is also required.

[0046] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A system for evaluating the effectiveness of conflict resolution, characterized in that: include: (a) Setting weights and scores: First, set the weights and scores of factors and levels in the system; (ii) Labeling: Label the factors and levels in the system cases, and feed the labeling results into the model for training; (iii) Deep learning model training: The labeled results are trained in the fused model, that is, the data is trained in the UIE large model; (iv) Predictive reasoning: After obtaining the prediction results through the deep learning model, the final performance prediction score is obtained through a custom algorithm.

2. The conflict resolution effectiveness evaluation system according to claim 1, characterized in that: The factors and levels refer to the diversion effectiveness factors and diversion effectiveness levels, respectively, and both can be queried, modified, and exported within the system.

3. The conflict resolution effectiveness evaluation system according to claim 2, characterized in that: The factors for improving the effectiveness of mediation include: the degree of achievement of demands, emotional satisfaction, time to resolve conflicts, frequency of mediation, and scope of administrative intervention, with corresponding weights of 20%, 30%, 20%, 15%, and 15%, respectively.

4. The conflict resolution effectiveness evaluation system according to claim 3, characterized in that: The guidance effectiveness levels are divided into: low, relatively low, average, relatively high, and high, with corresponding score ranges of 0-20, 21-40, 41-60, 61-80, and 81-100, respectively.

5. The conflict resolution effectiveness evaluation system according to claim 4, characterized in that: The annotation result refers to the annotated content of the dataset, which is input in text form.

6. The conflict resolution effectiveness evaluation system according to claim 5, characterized in that: The custom algorithm is based on the CNN convolutional neural network model and outputs a score and text format. The CNN convolutional neural network can extract data features and realize hierarchical representation and learning of features. In addition, different weights are set for the five dimensions of the mediation effectiveness: the degree of achievement of demands, emotional satisfaction, conflict resolution time, frequency of conflict mediation, and scope of administrative intervention.

7. The conflict resolution effectiveness evaluation system according to claim 6, characterized in that: The custom algorithm is mainly a fusion of a CNN convolutional neural network model and a weighted multiplication algorithm; the specific steps are as follows: ① Traverse the information of the five dimensions: The prediction data obtained through deep learning will contain information of different dimensions. It is necessary to first obtain the keywords of each dimension factor in the case. ② CNN Convolutional Neural Network Classification: Keywords for each dimension factor are input into a CNN convolutional neural network for classification, allowing the machine to automatically identify the performance level of the keyword; ③ Weighted multiplication algorithm: The effectiveness level and corresponding factor of the factors in different dimensions are automatically identified in the CNN convolutional neural network. The score is multiplied by the weight of the corresponding factor, and then the scores of different factors after being multiplied by the weight are added together to obtain the final total score. This score is the total effectiveness score of the case. ④ Determine the level within the interval: The total performance score will be determined within a pre-set interval to determine the performance level of that score.

Images