Training method for classifiers used in dispute resolution classification

Abstract

The present application discloses a training method for a classifier for dispute resolution classification based on weighted feature expression, which includes: obtaining a dispute data set, which is divided into multiple major categories and each major category contains multiple subcategories for representing major category indicators; converting multiple dispute data of each subcategory into feature vectors and performing one-dimensional convolution to obtain a subcategory dispute feature vector; grouping the subcategory dispute feature vectors according to major categories and fusing them to obtain multiple major category dispute feature vectors; performing one-dimensional convolution on the multiple major category dispute feature vectors and activating them with a Sigmoid function to obtain multiple major category dispute weighting coefficients; weighting the multiple major category dispute feature vectors with the multiple major category dispute weighting coefficients to obtain a dispute classification feature vector; classifying the dispute classification feature vector as label data of the corresponding case text feature vector through a classifier to obtain a classification loss function value; and updating the parameters of the classifier based on the classification loss function value.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a training method for a classifier for dispute resolution classification based on weighted feature expression, a method, system, and electronic device for classifying dispute resolution units. Background Art

[0002] To advance the dispute resolution functions of political and legal organs, we hope to comprehensively enhance the digital governance capabilities and level of the political and legal system through the digitization of internal operations and management, as well as the digitization of the performance of functions. Specifically, for different types of disputes, such as civil-to-criminal disputes, source-to-incident disputes, and court litigation disputes, we hope to be able to categorize and assign each type of dispute to different grassroots grids for handling, such as subdistricts, communities, social organizations, grassroots courts, and grassroots public security organs. This will allow more conflicts and disputes to be properly resolved in the budding stage, outside of litigation, and effectively achieve early prevention, early detection, and early resolution.

[0003] Therefore, it is expected that dispute data can be reasonably allocated based on the specific type of dispute and the relevant data obtained during the dispute resolution process, such as the occurrence rate of different types of disputes, the success rate of resolution, etc., so that the most reasonable unit can be assigned to handle different disputes.

[0004] Currently, deep learning and neural networks are widely used in fields such as computer vision, natural language processing, and speech signal processing. Furthermore, deep learning and neural networks have demonstrated capabilities approaching or even surpassing those of humans in areas such as machine vision, image classification, object detection, semantic segmentation, and text translation.

[0005] The development of deep learning and neural networks has provided new solutions and plans for the rational allocation of dispute resolution. Summary of the Invention

[0006] In order to solve the above technical problems, this application is proposed. Embodiments of the present application provide a method for training a classifier for dispute resolution classification based on weighted feature expression, a method for classifying dispute resolution units, a system, and an electronic device, which classify dispute data according to major categories, and then obtain subcategory index data of the dispute data under each major category. Furthermore, a subcategory data feature vector for representing the correlation features within the subcategory dispute data is obtained through one-dimensional convolution. The subcategory dispute feature vectors are then grouped according to major categories and fused to obtain multiple major category dispute feature vectors, so that each major category dispute feature vector can fully express the implicit correlation features between all subcategory data within the major category. Subsequently, multiple major category dispute feature vectors are subjected to one-dimensional convolution and activated with a Sigmoid function to obtain multiple major category dispute weighting coefficients. The correlation features between the major categories can be extracted to obtain a numerical expression of the weight of the features of each major category dispute in the overall classification. In this way, the multiple major category dispute feature vectors are weighted by the multiple major category dispute weighting coefficients to obtain a dispute classification feature vector, so that the dispute classification feature vector can fully express the correlation features of the dispute data at different levels after being divided into major categories and subcategories. Therefore, when the dispute classification feature vector is used as label data of the feature vector of the case text as dispute description data for classification, the classification accuracy of the classifier can be improved.

[0007] According to one aspect of the present application, a method for training a classifier for dispute resolution classification based on weighted feature expression is provided, which includes:

[0008] Obtaining a dispute dataset, wherein the dispute dataset is divided into a plurality of major categories, each major category includes a plurality of minor categories for representing major category indicators, and each minor category includes a plurality of dispute data based on a disposal unit;

[0009] Convert multiple dispute data of each subcategory into feature vectors and perform one-dimensional convolution to obtain the subcategory dispute feature vector;

[0010] Grouping the sub-category dispute feature vectors according to major categories and fusing them to obtain multiple major category dispute feature vectors;

[0011] Performing one-dimensional convolution on the feature vectors of the multiple major dispute categories and activating them with a Sigmoid function to obtain weighted coefficients of the multiple major dispute categories;

[0012] Weighting the plurality of major dispute feature vectors by the plurality of major dispute weighting coefficients to obtain a dispute classification feature vector;

[0013] Classifying the dispute classification feature vector as label data of the corresponding case text feature vector through a classifier to obtain a classification loss function value, the classifier comprising a deep neural network and a predetermined classification function, and the predetermined classification function using the disposal unit as a label; and

[0014] The parameters of the classifier are updated based on the classification loss function value.

[0015] In the above-mentioned training method of the classifier for dispute resolution classification based on weighted feature expression, multiple dispute data of each subcategory are converted into feature vectors and one-dimensional convolution is performed to obtain the subcategory dispute feature vector, including: obtaining the major category number, minor category number, minor category value and disposal unit number corresponding to each dispute data; normalizing the major category number, minor category number, minor category value and disposal unit number to obtain a feature vector in the form of a four-element array; and, performing one-dimensional convolution on the feature vector to obtain the subcategory dispute feature vector.

[0016] In the above-mentioned training method of the classifier for dispute resolution classification based on weighted feature expression, the small category dispute feature vectors are grouped according to large categories and fused to obtain multiple large category dispute feature vectors, including: grouping the small category dispute feature vectors according to large categories; and cascading the multiple small category dispute feature vectors in each group to obtain multiple large category dispute feature vectors.

[0017] In the above-mentioned training method of a classifier for dispute resolution classification based on weighted feature expression, the multiple major dispute feature vectors are subjected to one-dimensional convolution and activated with a Sigmoid function to obtain multiple major dispute weighting coefficients, including: calculating the average value or maximum value of each major dispute feature vector in the multiple major dispute feature vectors; constructing the multiple average values ​​or the maximum values ​​into a major numerical feature vector; and, performing one-dimensional convolution on the major numerical feature vector and activating it with a Sigmoid function to obtain multiple major dispute weighting coefficients.

[0018] In the above-mentioned method for training a classifier for dispute resolution classification based on weighted feature expression, the major categories and corresponding minor categories include:

[0019] Disputes that transfer from civil to criminal cases, including the number of cases, success rate of resolution, and rate of transfer from civil to criminal cases;

[0020] Visit the source of incident-related disputes, including the number of incidents and the success rate of resolution;

[0021] Court litigation disputes, including the number of cases and the success rate of resolution; and

[0022] Disputes related to collaborative case handling include the total number of transferred cases, the number of electronic bail changes, the number of dual-track transferred cases, the number of single-track transferred cases, and the arrest coordination rate.

[0023] According to another aspect of the present application, a method for classifying dispute resolution units is provided, comprising:

[0024] Obtaining dispute data to be handled, wherein the dispute data includes a case description;

[0025] Converting the case description into a case text feature vector;

[0026] The case text feature vector is input into the classifier trained by the training method of the classifier for dispute resolution classification based on weighted feature expression as described above to obtain a classification result, and the classification result is used to represent the disposal unit of the dispute data to be disposed.

[0027] According to another aspect of the present application, a training system for a classifier for dispute resolution classification based on weighted feature expression is provided, comprising:

[0028] a dispute data set acquisition unit, configured to obtain a dispute data set, wherein the dispute data set is divided into a plurality of major categories, each major category comprising a plurality of minor categories for representing major category indicators, and each minor category comprising a plurality of dispute data based on a disposal unit;

[0029] A sub-category dispute feature vector generating unit, configured to convert the multiple dispute data of each sub-category obtained by the dispute data set obtaining unit into a feature vector and perform one-dimensional convolution to obtain the sub-category dispute feature vector;

[0030] a major dispute feature vector generating unit, configured to group the minor dispute feature vectors obtained by the minor dispute feature vector generating unit according to major categories and fuse them to obtain a plurality of major dispute feature vectors;

[0031] a major dispute weighting coefficient generating unit, configured to perform one-dimensional convolution on the plurality of major dispute feature vectors obtained by the major dispute feature vector generating unit and activate the convolution with a Sigmoid function to obtain a plurality of major dispute weighting coefficients;

[0032] a dispute classification feature vector generating unit, configured to weight the plurality of major dispute feature vectors obtained by the major dispute feature vector generating unit using the plurality of major dispute weight coefficients obtained by the major dispute weight coefficient generating unit to obtain a dispute classification feature vector;

[0033] a classification loss function value calculation unit, configured to classify the dispute classification feature vector obtained by the dispute classification feature vector generation unit as label data of the corresponding case text feature vector through a classifier to obtain a classification loss function value, wherein the classifier includes a deep neural network and a predetermined classification function, and the predetermined classification function uses the disposal unit as a label; and

[0034] A parameter updating unit is used to update the parameters of the classifier based on the classification loss function value.

[0035] In the training system of the classifier for dispute resolution classification based on the above-mentioned weight-based feature expression, the small category dispute feature vector generation unit includes: a dispute data indicator acquisition subunit, used to obtain the major category number, minor category number, minor category value and disposal unit number corresponding to each dispute data; a four-element feature vector generation subunit, used to normalize the major category number, minor category number, minor category value and disposal unit number to obtain a feature vector in the form of a four-element array; and a one-dimensional convolution processing subunit, used to perform one-dimensional convolution on the feature vector to obtain a small category dispute feature vector.

[0036] In the training system of the classifier for dispute resolution classification based on weighted feature expression, the major dispute feature vector generation unit includes: a grouping subunit for grouping the minor dispute feature vectors according to major categories; and a cascading subunit for cascading multiple minor dispute feature vectors in each group to obtain multiple major dispute feature vectors.

[0037] In the training system of the classifier for dispute resolution classification based on weighted feature expression, the major dispute weight coefficient generation unit includes: a numerical calculation subunit, used to calculate the average value or maximum value of each major dispute feature vector in the multiple major dispute feature vectors; a major category numerical feature vector generation subunit, used to construct the multiple average values ​​or the maximum values ​​into a major category numerical feature vector; and a one-dimensional convolution activation subunit, used to perform one-dimensional convolution on the major category numerical feature vector and activate it with a Sigmoid function to obtain multiple major dispute weight coefficients.

[0038] In the above-mentioned training system of the classifier for dispute resolution classification based on weighted feature expression, the major categories and corresponding minor categories include:

[0039] Disputes that transfer from civil to criminal cases, including the number of cases, success rate of resolution, and rate of transfer from civil to criminal cases;

[0040] Visit the source of incident-related disputes, including the number of incidents and the success rate of resolution;

[0041] Court litigation disputes, including the number of cases and the success rate of resolution; and

[0042] Disputes related to collaborative case handling include the total number of transferred cases, the number of electronic bail changes, the number of dual-track transferred cases, the number of single-track transferred cases, and the arrest coordination rate.

[0043] According to another aspect of the present application, a dispute resolution unit classification system is provided, which includes:

[0044] a pending dispute data acquisition unit, configured to acquire pending dispute data, wherein the dispute data includes a case description;

[0045] a case text feature vector generating unit, configured to convert the case description obtained by the pending dispute data obtaining unit into a case text feature vector; and

[0046] A classification unit is used to input the case text feature vector obtained by the case text feature vector generation unit into a classifier trained by the training method of the classifier for dispute resolution classification based on weighted feature expression as described above to obtain a classification result, and the classification result is used to represent the disposal unit of the dispute data to be disposed of.

[0047] According to another aspect of the present application, an electronic device is provided, comprising: a processor; and a memory, in which computer program instructions are stored, and when the computer program instructions are executed by the processor, the processor executes the training method of a classifier for dispute resolution classification based on weight-based feature expression as described above, or the classification method of dispute resolution units.

[0048] According to another aspect of the present application, a computer-readable medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the processor executes the training method of a classifier for dispute resolution classification based on weight-based feature expression as described above, or the classification method of dispute resolution units.

[0049] The present application provides a training method for a classifier for dispute resolution classification based on weighted feature expression, a classification method for dispute resolution units, a system, and an electronic device. The method classifies dispute data according to major categories, and then obtains subcategory index data of the dispute data under each major category. Furthermore, a subcategory data feature vector for representing the correlation features within the subcategory dispute data is obtained through one-dimensional convolution. The subcategory dispute feature vectors are then grouped according to major categories and fused to obtain multiple major category dispute feature vectors, so that each major category dispute feature vector can fully express the implicit correlation features between all subcategory data within the major category. Subsequently, multiple major category dispute feature vectors are subjected to one-dimensional convolution and activated with a Sigmoid function to obtain multiple major category dispute weighting coefficients. The correlation features between the major categories can be extracted to obtain a numerical expression of the weight of the features of each major category dispute in the overall classification. In this way, the multiple major category dispute feature vectors are weighted by the multiple major category dispute weighting coefficients to obtain a dispute classification feature vector, so that the dispute classification feature vector can fully express the correlation features of the dispute data at different levels after being divided into major categories and minor categories. Therefore, when the dispute classification feature vector is used as label data of the feature vector of the case text as dispute description data for classification, the classification accuracy of the classifier can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0051] Figure 1 The flowchart of the method for training a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application is illustrated.

[0052] Figure 2 The diagram illustrates an architectural diagram of a method for training a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application.

[0053] Figure 3 The diagram illustrates a flowchart of a method for training a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application, in which multiple dispute data of each subcategory are converted into feature vectors and one-dimensional convolution is performed to obtain subcategory dispute feature vectors.

[0054] Figure 4The diagram illustrates a flowchart of a training method for a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application, in which the feature vectors of multiple major dispute categories are subjected to one-dimensional convolution and activated with a Sigmoid function to obtain weighted coefficients of multiple major dispute categories.

[0055] Figure 5 The figure illustrates a flow chart of a method for classifying dispute resolution units according to an embodiment of the present application.

[0056] Figure 6 The diagram illustrates a block diagram of a training system for a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application.

[0057] Figure 7 The diagram illustrates a block diagram of a sub-category dispute feature vector generating unit in a training system for a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application.

[0058] Figure 8 The diagram illustrates a block diagram of a unit for generating feature vectors of major categories of disputes in a training system for a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application.

[0059] Figure 9 The diagram illustrates a block diagram of a unit for generating weighted coefficients for major categories of disputes in a training system for a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application.

[0060] Figure 10 The figure illustrates a block diagram of a dispute resolution unit classification system according to an embodiment of the present application.

[0061] Figure 11 A block diagram of an electronic device according to an embodiment of the present application is illustrated. DETAILED DESCRIPTION

[0062] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0063] Application Overview

[0064] As mentioned above, in order to promote the dispute resolution functions of political and legal organs and comprehensively improve the digital governance capabilities and levels of the political and legal system through the digitalization of internal operation management and the digitalization of function performance, it is expected that dispute data can be reasonably allocated based on the specific types of disputes and relevant data obtained during the dispute resolution process, such as the occurrence rate of different types of disputes, the success rate of resolution, etc., so that the most reasonable units can be assigned to handle different disputes.

[0065] In recent years, the development of deep learning and neural networks has provided new solutions and plans for the rational allocation of dispute resolution.

[0066] Through research and development, the inventors of this application realized that the problem of reasonable allocation of dispute resolution is essentially a data-based classification problem. However, relying solely on deep neural networks to classify raw data fails to effectively extract the types used for specific classification, i.e., the effective features for the classification based on the disposal unit in this application, as the raw data cannot reflect the specific types of disputes and their more detailed indicators. This affects the classification results.

[0067] Therefore, in the technical solution of this application, dispute data is first classified according to major categories, for example, civil-to-criminal, source-to-incident, court litigation, street or community mediation, public security agency mediation, etc., and then sub-category indicator data for the dispute data under each major category is obtained, such as the number of occurrences, resolution success rate, etc. Furthermore, the applicant of this application discovered that after classification based on sub-category indicators, multiple pieces of dispute data within each sub-category can basically be represented in numerical form. While this facilitates calculations using mathematical models, the implicit associations between numerical values ​​are more difficult to mine than in other forms such as text and images.

[0068] Therefore, in the technical solution of the present application, first, a small category data feature vector for representing the correlation features within the small category dispute data is obtained through one-dimensional convolution, and then the small category dispute feature vectors are grouped according to the major category and fused to obtain multiple major category dispute feature vectors, so that each major category dispute feature vector can fully express the implicit correlation features between all the small category data within the major category. Then, by performing one-dimensional convolution on the multiple major category dispute feature vectors and activating them with a Sigmoid function to obtain multiple major category dispute weighting coefficients, the numerical expression of the weight of the features representing each major category dispute in the overall classification can be obtained by extracting the correlation features between the major categories. In this way, by weighting the multiple major category dispute feature vectors with multiple major category dispute weighting coefficients to obtain a dispute classification feature vector, the dispute classification feature vector can be made able to fully express the correlation features of the dispute data at different levels after being divided into major categories and minor categories.

[0069] Therefore, when the dispute classification feature vector is used as label data of the feature vector of the case text as dispute description data for classification, the accuracy of classification can be improved.

[0070] Based on this, the present application proposes a training method for a classifier for dispute resolution classification based on weighted feature expression, which includes: obtaining a dispute data set, wherein the dispute data set is divided into multiple major categories, and each major category contains multiple subcategories for representing major category indicators, and each subcategory contains multiple dispute data based on resolution units; converting the multiple dispute data of each subcategory into feature vectors and performing one-dimensional convolution to obtain a small category dispute feature vector; grouping the small category dispute feature vectors according to major categories and fusing them to obtain multiple major category dispute feature vectors; performing one-dimensional convolution on the multiple major category dispute feature vectors and activating them with a Sigmoid function to obtain multiple major category dispute weighting coefficients; weighting the multiple major category dispute feature vectors with the multiple major category dispute weighting coefficients to obtain a dispute classification feature vector; using the dispute classification feature vector as the label data of the corresponding case text feature vector to classify through a classifier to obtain a classification loss function value, the classifier including a deep neural network and a predetermined classification function, and the predetermined classification function is labeled with the resolution unit; and updating the parameters of the classifier based on the classification loss function value.

[0071] Furthermore, the present application provides a method for classifying disposal units for dispute resolution, which includes: obtaining dispute data to be resolved, the dispute data including a case description; converting the case description into a case text feature vector; and inputting the case text feature vector into a classifier trained by the training method of the classifier for dispute resolution classification based on weighted feature expression as described above to obtain a classification result, wherein the classification result is used to represent the disposal unit of the dispute data to be resolved.

[0072] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0073] Exemplary Methods

[0074] Figure 1 The figure shows a flow chart of a method for training a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application. Figure 1As shown, the training method of the classifier for dispute resolution classification based on weighted feature expression according to the embodiment of the present application includes: S110, obtaining a dispute data set, wherein the dispute data set is divided into multiple major categories, and each major category contains multiple minor categories for representing major category indicators, and each minor category contains multiple dispute data based on the resolution unit; S120, converting the multiple dispute data of each minor category into a feature vector and performing one-dimensional convolution to obtain a minor category dispute feature vector; S130, grouping the minor category dispute feature vectors according to major categories and fusing them to obtain multiple major category dispute feature vectors; S140, converging the multiple major category dispute feature vectors. The dispute feature vector is subjected to one-dimensional convolution and activated with a Sigmoid function to obtain a plurality of major dispute weighting coefficients; S150, the plurality of major dispute feature vectors are weighted with the plurality of major dispute weighting coefficients to obtain a dispute classification feature vector; S160, the dispute classification feature vector is used as label data of the corresponding case text feature vector to be classified by a classifier to obtain a classification loss function value, the classifier includes a deep neural network and a predetermined classification function, and the predetermined classification function is labeled with the disposal unit; and, S170, the parameters of the classifier are updated based on the classification loss function value.

[0075] Figure 2 The diagram shows a schematic diagram of the architecture of a method for training a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application. Figure 2 As shown, in the network architecture of the training method, first, the feature vectors (e.g., Figure 2 V0 as shown in the figure) with convolutional layers (e.g. Figure 2 Cl1) is used to perform a one-dimensional convolution to obtain the feature vector of small class disputes (e.g., Figure 2 Then, the sub-category dispute feature vectors are grouped according to the major categories and fused to obtain multiple major category dispute feature vectors (for example, Figure 2 Then, the plurality of major dispute feature vectors are transformed into a convolutional layer (e.g., Figure 2 Cl2) is convolved in one dimension and the sigmoid function (e.g., Figure 2 The circle S shown in FIG is activated to obtain a plurality of major dispute weighting coefficients; then, the plurality of major dispute feature vectors are weighted by the plurality of major dispute weighting coefficients to obtain a dispute classification feature vector (for example, Figure 2 Then, the dispute classification feature vector is used as the label data of the corresponding case text feature vector to be classified by a classifier to obtain a classification loss function value, and the classifier includes a deep neural network (for example, Figure 2DN as shown in ) and a predetermined classification function (e.g. Figure 2 ), and the predetermined classification function uses the treatment unit as a label; then, the parameters of the classifier are updated based on the classification loss function value.

[0076] In step S110, a dispute dataset is obtained. The dispute dataset is divided into multiple major categories, each of which contains multiple subcategories representing major category indicators, and each subcategory contains multiple pieces of dispute data based on the disposal unit. In other words, in this embodiment of the application, the dispute data is organized according to major categories and subcategories.

[0077] Here, in the embodiment of the present application, the major categories include civil-to-criminal disputes, source-to-incident disputes, court litigation disputes and collaborative case-handling disputes, and their corresponding subcategories are: the civil-to-criminal disputes, including the number of cases, the success rate of resolution, and the civil-to-criminal rate; the source-to-incident disputes, including the number of incidents and the success rate of resolution; the court litigation disputes, including the number of cases and the success rate of resolution; the collaborative case-handling disputes, including the total number of transferred cases, the number of electronic bail changes, the number of dual-track transferred cases, the number of single-track transferred cases, and the arrest coordination rate.

[0078] Of course, it should be understood that in the embodiments of the present application, the major categories and minor categories used to organize dispute data can be adjusted based on actual conditions, and this is not limited to the present application.

[0079] In step S120, the multiple dispute data items for each subcategory are converted into feature vectors and subjected to one-dimensional convolution to obtain subcategory dispute feature vectors. It should be understood that each major category of dispute data contains subcategory indicator data, such as the number of occurrences and success rate. Therefore, after classification based on subcategory indicators, the multiple dispute data items within each subcategory can essentially be represented in numerical form. While this facilitates calculations using mathematical models, the implicit connections between numerical values ​​are more difficult to mine than in other formats such as text and images.

[0080] Accordingly, in an embodiment of the present application, a small category data feature vector for representing the associated features within the small category dispute data is obtained by one-dimensional convolution. Specifically, in an embodiment of the present application, the process of converting multiple dispute data of each small category into a feature vector and performing one-dimensional convolution to obtain the small category dispute feature vector includes: first, obtaining the major category number, minor category number, minor category value and disposal unit number corresponding to each dispute data, for example, major category number, civil to criminal dispute, number 01, minor category number, resolution success rate, number 02, minor category value, resolution success rate, 50%, disposal unit number, grassroots court, number 05. In this way, each data is represented by a value.

[0081] Then, the major category number, minor category number, minor category value, and disposal unit number are normalized to obtain a feature vector in the form of a four-element array. It should be understood that the numerical standard can be unified to facilitate calculation. Next, the feature vector is convolved in one dimension to obtain a minor category dispute feature vector.

[0082] That is, in the embodiment of the present application, each dispute data is represented by a numerical value, and the array is subjected to one-dimensional convolution, which can fully explore the correlation between the numerical values ​​represented by each data item, thereby fully extracting the characteristic expressions that can reflect the major categories, minor categories, numerical values ​​and disposal units of the dispute.

[0083] Figure 3 The figure shows a flow chart of converting multiple dispute data of each subclass into feature vectors and performing one-dimensional convolution to obtain subclass dispute feature vectors in a training method of a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application. Figure 3 As shown, multiple dispute data of each subcategory are converted into feature vectors and one-dimensional convolution is performed to obtain the subcategory dispute feature vector, including: S210, obtaining the major category number, minor category number, minor category value and disposal unit number corresponding to each dispute data; S220, normalizing the major category number, minor category number, minor category value and disposal unit number to obtain a feature vector in the form of a four-element array; and, S230, performing one-dimensional convolution on the feature vector to obtain the subcategory dispute feature vector.

[0084] In step S130, the sub-category dispute feature vectors are grouped according to the major category and fused to obtain multiple major category dispute feature vectors. In other words, the sub-category dispute feature vectors are grouped according to the major category and fused to obtain multiple major category dispute feature vectors, so that each major category dispute feature vector can fully express the implicit association characteristics between the data of all sub-categories within the major category.

[0085] Specifically, in an embodiment of the present application, the process of grouping the small-category dispute feature vectors according to large categories and fusing them to obtain multiple large-category dispute feature vectors includes: first, grouping the small-category dispute feature vectors according to large categories, where the large categories include the collaborative case-handling disputes mentioned above, including the total number of transferred cases, the number of electronic bail changes, the number of dual-track transferred cases, the number of single-track transferred cases, and the arrest coordination rate; then, cascading the multiple small-category dispute feature vectors in each group to obtain multiple large-category dispute feature vectors.

[0086] It should be understood that concatenating features can leverage the extracted features without losing their representation. Furthermore, concatenation allows us to learn the associations between the bits of the label vector and the data vector, as well as the associations between the bits of the label vector itself and the bits of the data vector itself. All of these learned associations can be presented in a unified global domain.

[0087] In step S140, the multiple major dispute feature vectors are subjected to one-dimensional convolution and activated with a Sigmoid function to obtain multiple major dispute weighting coefficients. It should be understood that by subjecting the multiple major dispute feature vectors to one-dimensional convolution and activating with a Sigmoid function to obtain multiple major dispute weighting coefficients, it is possible to extract the correlation features between the major categories to obtain a numerical expression representing the weight of the features of each major dispute category within the overall classification.

[0088] Specifically, in an embodiment of the present application, the process of performing one-dimensional convolution on the multiple major dispute feature vectors and activating them with a Sigmoid function to obtain multiple major dispute weighting coefficients includes: first calculating the average value or maximum value of each major dispute feature vector in the multiple major dispute feature vectors; then, constructing the multiple average values ​​or the maximum values ​​into a major numerical feature vector; then, performing one-dimensional convolution on the major numerical feature vector and activating it with a Sigmoid function to obtain multiple major dispute weighting coefficients.

[0089] In particular, because the large-category dispute weighting coefficient is used to learn the weight expression of small-category data relative to large-category data through one-dimensional convolution, it is equivalent to learning the characteristic expression of the contribution of the local characteristics of each small-category data to the overall characteristics of the large-category data. Therefore, the local characteristics of each small-category data can be calculated as a whole. Moreover, the local characteristics can be based on the characteristic attributes, and the average value or the maximum value can be considered. For example, the value used to represent the occurrence rate can consider its average value attribute, while the value used to represent the resolution success rate can consider its maximum value attribute.

[0090] Figure 4 The figure shows a flow chart of a method for training a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application, wherein the plurality of major dispute feature vectors are subjected to one-dimensional convolution and activated with a Sigmoid function to obtain weighted coefficients of the plurality of major dispute categories. Figure 4As shown, the plurality of major dispute feature vectors are subjected to one-dimensional convolution and activated with a Sigmoid function to obtain a plurality of major dispute weighting coefficients, including: S310, calculating the average value or maximum value of each major dispute feature vector in the plurality of major dispute feature vectors; S320, constructing the plurality of average values ​​or the maximum values ​​into a major numerical feature vector; and, S330, performing one-dimensional convolution on the major numerical feature vector and activating with a Sigmoid function to obtain a plurality of major dispute weighting coefficients.

[0091] In step S150, the plurality of major dispute feature vectors are weighted by the plurality of major dispute weighting coefficients to obtain a dispute classification feature vector. In this way, the dispute classification feature vector can fully express the correlation characteristics of the dispute data at different levels after being divided into major categories and minor categories.

[0092] In step S160, the dispute classification feature vector is used as label data of the corresponding case text feature vector and classified through a classifier to obtain a classification loss function value. The classifier includes a deep neural network and a predetermined classification function, and the predetermined classification function uses the disposal unit as a label.

[0093] It should be understood that when the dispute classification feature vector is used as label data of the feature vector of the case text as dispute description data for classification, the classification accuracy of the classifier can be improved.

[0094] It is worth mentioning that in the classifier, the predetermined classification function can be a Softmax classification function, wherein the predetermined classification function is labeled with the disposal unit, and the disposal unit includes: streets, communities, social organizations, grassroots courts, grassroots public security organs, etc.

[0095] In summary, a training method for a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application is clarified, which classifies dispute data according to major categories, and then obtains subcategory index data of the dispute data under each major category, and then obtains subcategory data feature vectors for representing the correlation features within the subcategory dispute data through one-dimensional convolution, and then groups the subcategory dispute feature vectors according to major categories and fuses them to obtain multiple major category dispute feature vectors, so that each major category dispute feature vector can fully express the implicit correlation features between all subcategory data within the major category; then, by performing one-dimensional convolution on the multiple major category dispute feature vectors and activating them with a Sigmoid function to obtain multiple major category dispute weighting coefficients, the correlation features between the major categories can be extracted to obtain a numerical expression of the weight of the features for representing each major category dispute in the overall classification, and in this way, by weighting the multiple major category dispute feature vectors with multiple major category dispute weighting coefficients to obtain a dispute classification feature vector, the dispute classification feature vector can be made to fully express the correlation features of the dispute data at different levels after being divided into major categories and minor categories. Therefore, when the dispute classification feature vector is used as label data of the feature vector of the case text as dispute description data for classification, the classification accuracy of the classifier can be improved.

[0096] Furthermore, after a classifier is trained using the above-described method for training a classifier for dispute resolution classification based on weighted feature expression, the classifier can be used to classify dispute data to be resolved into resolution units.

[0097] Figure 5 FIG2 is a flowchart of a method for classifying dispute resolution units according to an embodiment of the present application. Figure 5 As shown, the method for classifying disposal units for dispute resolution according to an embodiment of the present application includes: S410, obtaining dispute data to be disposed, wherein the dispute data includes a case description; S420, converting the case description into a case text feature vector; and, S430, inputting the case text feature vector into a classifier trained by the training method of the classifier for dispute resolution classification based on weighted feature expression as described above to obtain a classification result, wherein the classification result is used to represent the disposal unit of the dispute data to be disposed.

[0098] It is worth mentioning that when classifying the dispute data to be processed, part of the data in the dispute data to be processed can be extracted as labels. For example, the major category labels can be extracted from the dispute data to be processed as labels, and then classified together with the case text features. In this way, the classification difficulty can be reduced and the classification efficiency and accuracy can be improved.

[0099] Exemplary Systems

[0100] Figure 6 The diagram illustrates a block diagram of a training system for a classifier for dispute resolution classification based on weighted feature expression according to an embodiment of the present application.

[0101] like Figure 6 As shown, according to the embodiment of the present application, a training system 600 for a classifier for dispute resolution classification based on weighted feature expression includes: a dispute data set acquisition unit 610, for obtaining a dispute data set, wherein the dispute data set is divided into multiple major categories, and each major category includes multiple minor categories for representing major category indicators, and each minor category includes multiple dispute data based on a resolution unit; a minor category dispute feature vector generation unit 620, for converting multiple dispute data of each minor category obtained by the dispute data set acquisition unit 610 into a feature vector and performing one-dimensional convolution to obtain a minor category dispute feature vector; a major category dispute feature vector generation unit 630, for grouping the minor category dispute feature vectors obtained by the minor category dispute feature vector generation unit 620 according to major categories and fusing them to obtain multiple major category dispute feature vectors; a major category dispute weighting coefficient generation unit 640, for converting the minor category dispute feature vectors obtained by the major category dispute feature vector generation unit 630 into a feature vector; The multiple major dispute feature vectors are subjected to one-dimensional convolution and activated with a Sigmoid function to obtain multiple major dispute weighting coefficients; a dispute classification feature vector generating unit 650 is used to weight the multiple major dispute feature vectors obtained by the major dispute weighting coefficient generating unit 640 with the multiple major dispute weighting coefficients to obtain a dispute classification feature vector; a classification loss function value calculating unit 660 is used to classify the dispute classification feature vector obtained by the dispute classification feature vector generating unit 650 as the label data of the corresponding case text feature vector through a classifier to obtain a classification loss function value, the classifier includes a deep neural network and a predetermined classification function, and the predetermined classification function uses the disposal unit as a label; and a parameter updating unit 670 is used to update the parameters of the classifier based on the classification loss function value.

[0102] In one example, in the above training system 600, as Figure 7 As shown, the small category dispute feature vector generation unit 620 includes: a dispute data index acquisition subunit 621, which is used to obtain the major category number, minor category number, minor category value and disposal unit number corresponding to each dispute data; a four-element feature vector generation subunit 622, which is used to normalize the major category number, minor category number, minor category value and disposal unit number to obtain a feature vector in the form of a four-element array; and a one-dimensional convolution processing subunit 623, which is used to perform one-dimensional convolution on the feature vector to obtain a small category dispute feature vector.

[0103] In one example, in the above training system 600, as Figure 8 As shown, the major category dispute feature vector generation unit 630 includes: a grouping subunit 631, used to group the minor category dispute feature vectors according to major categories; and a cascading subunit 632, used to cascade multiple minor category dispute feature vectors in each group to obtain multiple major category dispute feature vectors.

[0104] In one example, in the above training system 600, as Figure 9 As shown, the major dispute weighting coefficient generating unit 640 includes: a numerical calculation subunit 641, used to calculate the average value or maximum value of each major dispute feature vector in the multiple major dispute feature vectors; a major category numerical feature vector generating subunit 642, used to construct the multiple average values ​​or the maximum values ​​into a major category numerical feature vector; and a one-dimensional convolution activation subunit 643, used to perform one-dimensional convolution on the major category numerical feature vector and activate it with a Sigmoid function to obtain multiple major dispute weighting coefficients.

[0105] In one example, in the above training system 600, the major categories and corresponding minor categories include:

[0106] Disputes that transfer from civil to criminal cases, including the number of cases, success rate of resolution, and rate of transfer from civil to criminal cases;

[0107] Visit the source of incident-related disputes, including the number of incidents and the success rate of resolution;

[0108] Court litigation disputes, including the number of cases and the success rate of resolution; and

[0109] Disputes related to collaborative case handling include the total number of transferred cases, the number of electronic bail changes, the number of dual-track transferred cases, the number of single-track transferred cases, and the arrest coordination rate.

[0110] Here, those skilled in the art will appreciate that the specific functions and operations of the various units and modules in the training system 600 have been described above with reference to FIG. Figures 1 to 4 The training method of the classifier for dispute resolution classification based on weight-based feature expression has been introduced in detail, and therefore, its repeated description will be omitted.

[0111] As described above, the training system 600 according to an embodiment of the present application can be implemented in various terminal devices, such as a server classified by a dispute data handling unit. In one example, the training system 600 according to an embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the training system 600 can be a software module in the operating system of the terminal device, or an application developed specifically for the terminal device; of course, the training system 600 can also be one of the terminal device's many hardware modules.

[0112] Alternatively, in another example, the training system 600 and the terminal device may be separate devices, and the training system 600 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0113] According to another aspect of the present application, a dispute resolution unit classification system is also provided.

[0114] Figure 10 FIG2 is a block diagram of a dispute resolution unit classification system according to an embodiment of the present application. Figure 10 As shown, the disposal unit classification system 700 for dispute resolution according to an embodiment of the present application includes: a pending dispute data acquisition unit, used to acquire pending dispute data, wherein the dispute data includes a case description; a case text feature vector generation unit, used to convert the case description obtained by the pending dispute data acquisition unit into a case text feature vector; and a classification unit, used to input the case text feature vector obtained by the case text feature vector generation unit into a classifier trained by the training method of the classifier for dispute resolution classification based on weighted feature expression as described above to obtain a classification result, wherein the classification result is used to represent the disposal unit of the pending dispute data.

[0115] Here, those skilled in the art will appreciate that the specific functions and operations of the various units and modules in the classification system 700 have been described in detail above. Figure 5 The description of the disposal unit classification method for dispute resolution has been introduced in detail, and therefore, its repeated description will be omitted.

[0116] As described above, the classification system 700 according to an embodiment of the present application can be implemented in various terminal devices, such as a server for classifying dispute data handling units. In one example, the classification system 700 according to an embodiment of the present application can be integrated into a terminal device as a software module and / or a hardware module. For example, the classification system 700 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the classification system 700 can also be one of the many hardware modules of the terminal device.

[0117] Alternatively, in another example, the classification system 700 and the terminal device may also be separate devices, and the classification system 700 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0118] Exemplary electronic devices

[0119] Below, reference Figure 11 To describe the electronic device according to the embodiment of the present application.

[0120] Figure 11 A block diagram of an electronic device according to an embodiment of the present application is illustrated.

[0121] like Figure 11 As shown, the electronic device 10 includes one or more processors 11 and a memory 12 .

[0122] The processor 11 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.

[0123] The memory 12 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), a hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 11 may execute the program instructions to implement the training of the classifier for dispute resolution classification based on weighted feature expression of the various embodiments of the present application described above, or the functions of the dispute resolution unit classification method and / or other desired functions. Various contents such as dispute data may also be stored in the computer-readable storage medium.

[0124] In one example, the electronic device 10 may further include an input device 13 and an output device 14 , and these components are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0125] The input device 13 may include, for example, a keyboard, a mouse, and the like.

[0126] The output device 14 can output various information to the outside, including classification results, etc. The output device 14 can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, and the like.

[0127] Of course, to simplify, Figure 11 Only some of the components related to the present application in the electronic device 10 are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, the electronic device 10 may further include any other appropriate components according to specific application scenarios.

[0128] Exemplary computer program products and computer-readable storage media

[0129] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to perform the training of a classifier for dispute resolution classification based on weight-based feature expression according to various embodiments of the present application, or the steps in the functions of a method for classifying dispute resolution units, as described in the above "Exemplary Method" section of this specification.

[0130] The computer program product may be written in any combination of one or more programming languages ​​to implement the program code for performing the operations of the embodiments of the present application, including object-oriented programming languages ​​such as Java, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0131] In addition, an embodiment of the present application may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, causes the processor to execute the training of a classifier for dispute resolution classification based on weight-based feature expression according to various embodiments of the present application, or the steps in the functions of a method for classifying dispute resolution units, as described in the above "Exemplary Method" section of this specification.

[0132] The computer-readable storage medium may be any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may include, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0133] The basic principles of the present application have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in this application are merely illustrative and not restrictive, and it should not be assumed that these advantages, strengths, and effects are required of each embodiment of this application. In addition, the specific details disclosed above are merely illustrative and facilitating understanding, and are not restrictive. The above details do not limit this application to necessarily being implemented using the above specific details.

[0134] The block diagrams of the devices, devices, equipment, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, devices, equipment, and systems can be connected, arranged, or configured in any manner. Words such as "include," "comprise," "have," and the like are open-ended words, meaning "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or" and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0135] It should also be noted that in the apparatus, device, and method of the present application, each component or each step can be decomposed and / or recombined, and such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0136] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0137] The above description has been provided for the purpose of illustration and description. Furthermore, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for classifying dispute resolution units, characterized in that: include: Obtaining dispute data to be handled, wherein the dispute data includes a case description; Converting the case description into a case text feature vector; Inputting the case text feature vector into a classifier trained by a training method for dispute resolution classification based on weighted feature expression to obtain a classification result, wherein the classification result is used to represent a resolution unit of the dispute data to be resolved; The steps of the method for training a classifier for dispute resolution classification based on weighted feature expression include: Obtaining a dispute dataset, wherein the dispute dataset is divided into a plurality of major categories, each major category includes a plurality of minor categories for representing major category indicators, and each minor category includes a plurality of dispute data based on a disposal unit; Convert multiple dispute data of each subcategory into feature vectors and perform one-dimensional convolution to obtain the subcategory dispute feature vector; Grouping the sub-category dispute feature vectors according to the major categories and fusing them to obtain multiple major category dispute feature vectors; Perform one-dimensional convolution on the feature vectors of the multiple major dispute categories and activate them with a Sigmoid function to obtain weight coefficients of the multiple major dispute categories; Weighting the plurality of major dispute feature vectors by the plurality of major dispute weighting coefficients to obtain a dispute classification feature vector; Classifying the dispute classification feature vector as label data of the corresponding case text feature vector through a classifier to obtain a classification loss function value, the classifier comprising a deep neural network and a predetermined classification function, and the predetermined classification function using the disposal unit as a label; and Updating parameters of the classifier based on the classification loss function value; Among them, multiple dispute data of each subcategory are converted into feature vectors and one-dimensional convolution is performed to obtain the subcategory dispute feature vector, including: Obtain the major category number, minor category number, minor category value and handling unit number corresponding to each dispute data; Normalizing the major category number, minor category number, minor category value, and disposal unit number to obtain a feature vector in a quaternion array form; and Performing one-dimensional convolution on the feature vector to obtain a small class dispute feature vector; The plurality of major dispute feature vectors are subjected to one-dimensional convolution and activated with a Sigmoid function to obtain a plurality of major dispute weighting coefficients, including: Calculating an average value or a maximum value of each major dispute feature vector among the multiple major dispute feature vectors; Constructing the average value or the maximum value of each major category dispute feature vector into a major category numerical feature vector; and One-dimensional convolution is performed on the large-category numerical feature vector and activated with a Sigmoid function to obtain a plurality of large-category dispute weighting coefficients.

2. The method for classifying dispute resolution units according to claim 1, wherein: The sub-category dispute feature vectors are grouped according to the major categories and fused to obtain multiple major category dispute feature vectors, including: Grouping the sub-category dispute feature vectors into major categories; and Multiple small-category dispute feature vectors in each group are concatenated to obtain multiple large-category dispute feature vectors.

3. The method for classifying dispute resolution units according to any one of claims 1 to 2, wherein: The major categories and corresponding subcategories include: Disputes that transfer from civil to criminal cases, including the number of cases, the success rate of resolution, and the rate of transfer from civil to criminal cases; Visit the source of incident-related disputes, including the number of incidents and the success rate of resolution; Court litigation disputes, including the number of cases and the success rate of resolution; and Disputes related to collaborative case handling include the total number of transferred cases, the number of electronic bail changes, the number of dual-track transferred cases, the number of single-track transferred cases, and the arrest coordination rate.

4. A dispute resolution unit classification system, characterized in that: include: a pending dispute data acquisition unit, configured to acquire pending dispute data, wherein the dispute data includes a case description; a case text feature vector generating unit, configured to convert the case description obtained by the pending dispute data obtaining unit into a case text feature vector; and a classification unit, configured to input the case text feature vector obtained by the case text feature vector generation unit into a classifier trained by a training method for a classifier for dispute resolution classification based on weighted feature expression to obtain a classification result, wherein the classification result is used to represent a resolution unit of the dispute data to be resolved; The steps of the method for training a classifier for dispute resolution classification based on weighted feature expression include: Obtaining a dispute dataset, wherein the dispute dataset is divided into a plurality of major categories, each major category includes a plurality of minor categories for representing major category indicators, and each minor category includes a plurality of dispute data based on a disposal unit; Convert multiple dispute data of each subcategory into feature vectors and perform one-dimensional convolution to obtain the subcategory dispute feature vector; Grouping the sub-category dispute feature vectors according to the major categories and fusing them to obtain multiple major category dispute feature vectors; Perform one-dimensional convolution on the feature vectors of the multiple major dispute categories and activate them with a Sigmoid function to obtain weight coefficients of the multiple major dispute categories; Weighting the plurality of major dispute feature vectors by the plurality of major dispute weighting coefficients to obtain a dispute classification feature vector; Classifying the dispute classification feature vector as label data of the corresponding case text feature vector through a classifier to obtain a classification loss function value, the classifier comprising a deep neural network and a predetermined classification function, and the predetermined classification function using the disposal unit as a label; and Updating parameters of the classifier based on the classification loss function value; Among them, multiple dispute data of each subcategory are converted into feature vectors and one-dimensional convolution is performed to obtain the subcategory dispute feature vector, including: Obtain the major category number, minor category number, minor category value and handling unit number corresponding to each dispute data; Normalizing the major category number, minor category number, minor category value, and disposal unit number to obtain a feature vector in a quaternion array form; and Performing one-dimensional convolution on the feature vector to obtain a small class dispute feature vector; The plurality of major dispute feature vectors are subjected to one-dimensional convolution and activated with a Sigmoid function to obtain a plurality of major dispute weighting coefficients, including: Calculating an average value or a maximum value of each major dispute feature vector among the multiple major dispute feature vectors; Constructing the average value or the maximum value of each major category dispute feature vector into a major category numerical feature vector; and One-dimensional convolution is performed on the large-category numerical feature vector and activated with a Sigmoid function to obtain a plurality of large-category dispute weighting coefficients.

5. The dispute resolution unit classification system according to claim 4, wherein: The sub-category dispute feature vectors are grouped according to the major categories and fused to obtain multiple major category dispute feature vectors, including: Grouping the sub-category dispute feature vectors into major categories; and Multiple small-category dispute feature vectors in each group are concatenated to obtain multiple large-category dispute feature vectors.

6. The dispute resolution unit classification system according to any one of claims 4 to 5, wherein: The major categories and corresponding subcategories include: Disputes that transfer from civil to criminal cases, including the number of cases, the success rate of resolution, and the rate of transfer from civil to criminal cases; Visit the source of incident-related disputes, including the number of incidents and the success rate of resolution; Court litigation disputes, including the number of cases and the success rate of resolution; and Disputes related to collaborative case handling include the total number of transferred cases, the number of electronic bail changes, the number of dual-track transferred cases, the number of single-track transferred cases, and the arrest coordination rate.

Images