Method, device, computer equipment and storage medium for processing judicial documents

Through machine learning models, the judgment documents are analyzed, the matching degree is detected and the evaluation results are generated, which solves the problem of low accuracy in criminal cases quality review in traditional technology, realizes a comprehensive evaluation of judicial documents, and improves the evaluation efficiency.

CN112561219BActive Publication Date: 2024-11-19BEIJING GRIDSUM TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN201910911885.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2019-09-25
Publication Date
2024-11-19
Estimated Expiration
2039-09-25

AI Technical Summary

Technical Problem

When the court evaluates the quality of criminal cases, the accuracy rate in traditional technology is low, making it difficult to effectively evaluate the quality of the case, resulting in inefficiency.

Method used

The machine learning model is used to analyze the judicial documents, and the evaluation results are generated through matching degree detection, including the evaluation of the conviction deviation, the type of punishment deviation and sentencing difference, and then the evaluation results of the judicial documents are generated, and the evaluation process is saved and indexed is established using the graphic data structure.

Benefits of technology

The accuracy and efficiency of comprehensive review of the crimes, types of sentences and sentences of judgment documents has been improved, and the court has assisted the court in accurately determining the quality of the case and improving the efficiency of the review.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112561219B_ABST
    Figure CN112561219B_ABST
Patent Text Reader

Abstract

The present application relates to a method, device, computer equipment and storage medium for processing judgment documents, which obtains the judgment document to be evaluated, inputs the judgment document to be evaluated into a machine learning model to extract the case features of the judgment document to be evaluated, and outputs the predicted judgment crime text and predicted judgment type text corresponding to the judgment document to be evaluated, thereby detecting the matching degree between the predicted judgment crime text and the predicted judgment type text and the actual judgment crime text and the actual judgment type text in the judgment document to be evaluated, so as to determine whether there is a deviation in conviction or a deviation in the type of punishment. The review of the judgment crime text and the judgment type text in the judgment document to be evaluated is realized, which can not only assist the trial management department of the court to accurately and quickly determine the quality of the case, but also improve the efficiency of the trial management department of the court in reviewing criminal cases.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, apparatus, computer equipment and storage medium for processing judicial documents. Background Art

[0002] The quality review of criminal cases is an important task carried out by judicial organs such as courts and procuratorates in order to strengthen internal supervision and manage case quality. It is also an important way to achieve judicial justice.

[0003] At present, the trial management department of the court will evaluate the quality of cases and judges through the quality review of criminal cases. At the same time, it will collect statistics on the quality of case handling by case handlers year by year, and establish a law enforcement and case handling file for full-time judges to provide reference for their work performance, promotion and job exchange. The review results of criminal case quality will also be provided to the political department as a reference for annual individual assessment and evaluation.

[0004] However, in traditional technology, the trial management department of the court mainly evaluates the verdicts of criminal cases by judging whether there is deviation in sentencing, which has the technical problem of low accuracy. Summary of the invention

[0005] Based on this, it is necessary to provide a method, device, computer equipment and storage medium for processing judicial documents to address the technical problem of inefficiency in case quality assessment in traditional technologies.

[0006] A method for processing a judgment document, the method comprising:

[0007] Obtaining the judgment document to be evaluated, wherein the judgment document to be evaluated includes the text of the actual crime and the text of the actual sentence;

[0008] Input the judgment document to be evaluated into the machine learning model for analysis to obtain the predicted judgment result corresponding to the judgment document to be evaluated, wherein the predicted judgment result includes the predicted judgment crime text and the predicted judgment type text;

[0009] Detecting a first degree of match between the actual judgment crime text and the predicted judgment crime text;

[0010] When the first matching degree is a complete match, detecting a second matching degree between the actual sentence text and the predicted sentence text;

[0011] Based on the detection result of the second matching degree, a first evaluation result of the judgment document to be evaluated is generated.

[0012] In one embodiment, the method further comprises:

[0013] When the first matching degree is an incomplete match, a determination result is generated that the judgment document to be evaluated is a conviction deviation document.

[0014] In one embodiment, the judgment document to be evaluated further includes an actual sentence value, and the predicted judgment result further includes a predicted sentence value; and generating a first evaluation result of the judgment document to be evaluated based on the detection result of the second matching degree includes:

[0015] When the second matching degree is an incomplete match, a determination result is generated that the judgment document to be evaluated is a document of deviation from the criminal type;

[0016] When the second matching degree is a complete match, detecting a sentencing difference value between the actual sentence value and the predicted sentence value;

[0017] Based on the detection result of the sentencing difference value, a second evaluation result of the judgment document to be evaluated is generated.

[0018] In one embodiment, a safety threshold is set between the actual sentence and the predicted sentence; and generating a second evaluation result of the judgment document to be evaluated based on the detection result of the sentencing difference value includes:

[0019] If the sentencing difference value is less than or equal to the safety threshold, a determination result is generated that the judgment document to be evaluated is a non-deviation document;

[0020] If the sentencing difference value is greater than the safety threshold, a determination result is generated that the judgment document to be evaluated belongs to the sentence deviation document.

[0021] In one embodiment, the safety threshold includes a first safety threshold, and the sentencing difference value includes an absolute difference value; if the sentencing difference value is greater than the safety threshold, generating a determination result that the judgment document to be evaluated belongs to the sentence deviation document, includes:

[0022] If the absolute difference value is greater than the first safety threshold, and the actual sentence value is greater than the predicted sentence value, a determination result is generated that the judgment document to be evaluated is a document with excessively severe sentence deviation;

[0023] If the absolute difference value is greater than the first safety threshold, and the actual sentence value is less than the predicted sentence value, then a determination result is generated that the judgment document to be evaluated is a deviation from the document with too light a sentence.

[0024] In one embodiment, the safety threshold includes a second safety threshold, and the sentencing difference value includes a relative difference value; if the sentencing difference value is greater than the safety threshold, generating a determination result that the judgment document to be evaluated belongs to the sentence deviation document, includes:

[0025] If the relative difference value is greater than the second safety threshold, and the actual sentence value is greater than the predicted sentence value, a determination result is generated that the judgment document to be evaluated is a document with excessively severe sentence deviation;

[0026] If the relative difference value is greater than the second safety threshold, and the actual sentence value is less than the predicted sentence value, then a determination result is generated that the judgment document to be evaluated is a deviation document with too light a sentence.

[0027] In one embodiment, the method further comprises:

[0028] The review process of the non-deviation document, the conviction deviation document, the punishment type deviation document, and the sentence deviation document is saved in a case review database in the form of a graphic data structure; and an index is established based on the trial judge, the crime sentenced, the punishment type sentenced, and the sentence sentenced.

[0029] A device for processing judicial documents, the device comprising:

[0030] A judgment document acquisition module is used to acquire the judgment document to be evaluated, wherein the judgment document to be evaluated includes the text of the actual crime and the text of the actual sentence;

[0031] A judgment prediction module is used to input the judgment document to be evaluated into a machine learning model for analysis to obtain a predicted judgment result corresponding to the judgment document to be evaluated, wherein the predicted judgment result includes a predicted judgment crime text and a predicted judgment type text;

[0032] A first detection module, used to detect a first matching degree between the actual judgment crime text and the predicted judgment crime text;

[0033] A second detection module is used to detect a second matching degree between the actual sentence text and the predicted sentence text when the first matching degree is a complete match;

[0034] An evaluation result generating module is used to generate a first evaluation result of the judgment document to be evaluated based on the detection result of the second matching degree.

[0035] A computer device comprises at least one processor, and at least one memory and a bus connected to the processor; wherein the processor and the memory communicate with each other via the bus; the processor is used to call program instructions in the memory to execute the steps of the method for processing judicial documents described in any of the above embodiments.

[0036] A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the method for processing judicial documents described in any of the above embodiments are implemented.

[0037] The above-mentioned judgment document processing method, device, computer equipment and storage medium obtain the judgment document to be evaluated, input the judgment document to be evaluated into the machine learning model to extract the case features of the judgment document to be evaluated, and output the predicted judgment crime text and predicted judgment type text corresponding to the judgment document to be evaluated, so as to detect the matching degree between the predicted judgment crime text and the predicted judgment type text and the actual judgment crime text and the actual judgment type text in the judgment document to be evaluated, so as to determine whether there is a deviation in conviction or a deviation in punishment. A comprehensive review of the judgment crime text and the judgment type text in the judgment document to be evaluated is achieved, which can assist the trial management department of the court to accurately and quickly determine the quality of the case. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 An application environment diagram of a method for processing judicial documents in one embodiment;

[0039] Figure 2 1. It is a flowchart of a method for processing a judgment document in one embodiment;

[0040] Figure 3 1. It is a flowchart of a method for processing a judgment document in one embodiment;

[0041] Figure 4 1. It is a flowchart of a method for processing a judgment document in one embodiment;

[0042] Figure 5 1. It is a flowchart of a method for processing a judgment document in one embodiment;

[0043] Figure 6 1. It is a flowchart of a method for processing a judgment document in one embodiment;

[0044] Figure 7 is a structural block diagram of a device for processing judicial documents in one embodiment;

[0045] Figure 8 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0047] This application provides a method for processing judicial documents, which can be applied to Figure 1 In the application environment shown. Among them, the terminal 110 exchanges data with the server 120 through the network. The terminal 110 can be used to receive the case delineation conditions of the judgment document to be evaluated set by the user, and the user triggers the corresponding button, and the terminal 110 initiates an evaluation instruction for the judgment document to be evaluated to the server 120, and the evaluation instruction carries the case delineation conditions of the judgment document to be evaluated. The server 120 receives the evaluation instruction of the judgment document to be evaluated through the terminal 110, and the server 120 obtains the case delineation conditions of the judgment document to be evaluated. The case delineation conditions include at least one of the geographical scope, the crime, the trial court, and the judgment date, and searches in the criminal case database according to the case delineation conditions to obtain the judgment document to be evaluated; the judgment document to be evaluated includes the actual judgment crime text and the actual judgment sentence text. Next, the obtained judgment document to be evaluated is input into the machine learning model, and the case characteristics of the judgment document to be evaluated are analyzed by the machine learning model to obtain the predicted judgment result corresponding to the judgment document to be evaluated, and the predicted judgment result includes the predicted judgment crime text and the predicted judgment sentence text; the first matching degree between the actual judgment crime text and the predicted judgment crime text is detected; when the first matching degree is a complete match, the second matching degree between the actual judgment sentence text and the predicted judgment sentence text is detected; according to the detection result of the second matching degree, the first evaluation result of the judgment document to be evaluated is generated. Among them, the terminal 110 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers and portable wearable devices, and the server 120 can be implemented by an independent server or a server cluster composed of multiple servers.

[0048] In one embodiment, Figure 2 As shown, a method for processing judicial documents is provided, and the method is applied to Figure 1 The following steps are taken as an example to illustrate the server:

[0049] S210. Obtain the judgment document to be evaluated.

[0050] Among them, the judgment document in this embodiment refers to the process and results of the People's Court hearing a criminal case, and it is the carrier of the results of the criminal case handling. The judgment document includes a judgment result section, which includes the crime and type of sentence sentenced by the trial judge to the parties in the criminal case. The crime refers to the name of a specific crime stipulated by the law, which is a high-level summary of the essential characteristics and main characteristics of the crime. The type of sentence is the abbreviation of the type of punishment, including control, detention, fixed-term imprisonment, life imprisonment and death penalty. The judgment document to be evaluated refers to the judgment document that needs to be evaluated for the quality of criminal case sentencing, and the judgment document to be evaluated includes the actual judgment crime text and the actual judgment type text.

[0051] Specifically, in order to review the quality of criminal cases, the server may store judgment documents to be evaluated locally in advance, and then obtain the judgment documents to be evaluated from the server locally, or obtain the judgment documents to be evaluated from a server that is communicatively connected to the server.

[0052] S220. Input the judgment document to be evaluated into the machine learning model for analysis to obtain the predicted judgment result corresponding to the judgment document to be evaluated.

[0053] Among them, the full name of machine learning in English is Machine Learning, abbreviated as ML. The machine learning model can have specific capabilities through sample learning. The machine learning model can adopt a neural network model, etc. Neural network models include convolutional neural networks, recurrent neural networks, etc. In this embodiment, the machine learning model is a model with predictive capabilities obtained through pre-training. The training process of the machine learning model may include: a computer device obtains a sample set of criminal case judgment documents, annotates the criminal case judgment documents with the crime and the type of sentence to obtain labels for training, and trains the machine learning model using the sample set and the label to obtain a machine learning model for predicting the judgment result. And the predicted judgment result includes the predicted judgment crime text and the predicted judgment type text.

[0054] Specifically, the judicial document to be evaluated is input into the machine learning model, and the machine learning model is used to analyze the case characteristics corresponding to the judicial document to be evaluated. Predictions are made based on the case characteristics obtained by the analysis. The machine learning model outputs the predicted judgment result corresponding to the judicial document to be evaluated. The predicted judgment result includes the predicted crime text and the predicted sentence type text.

[0055] S230, detecting a first degree of match between the actual judgment crime text and the predicted judgment crime text.

[0056] Specifically, the judgment result segment of the judgment document to be evaluated includes the actual judgment charge text, and the machine learning model is used to output the predicted judgment charge text corresponding to the judgment document to be evaluated. The actual judgment charge text is compared with the predicted judgment charge text, and the first matching degree between the actual judgment charge text and the predicted judgment charge text is detected to determine whether the two are consistent. If the first matching degree is a complete match, that is, the two are consistent, it means that there is no conviction deviation between the actual judgment charge and the predicted judgment charge. If the two are inconsistent, it means that there is a conviction deviation between the actual judgment charge and the predicted judgment charge.

[0057] S240: When the first matching degree is a complete match, detecting a second matching degree between the actual sentence text and the predicted sentence text;

[0058] S250. Generate a first evaluation result of the judgment document to be evaluated based on the detection result of the second matching degree.

[0059] Among them, since the judgment result segment includes not only the text of the actual judgment crime, but also the text of the actual judgment sentence, in order to comprehensively evaluate the quality of criminal cases, if the actual judgment crime text completely matches the predicted judgment crime text, and it is determined that there is no conviction deviation between the actual judgment crime and the predicted judgment crime, it is necessary to further evaluate the judgment document to be evaluated. Specifically, if there is no conviction deviation between the actual judgment crime and the predicted judgment crime, the actual judgment sentence text in the judgment document to be evaluated is compared with the predicted judgment sentence text output by the machine learning model to detect the second matching degree between the actual judgment sentence text and the predicted judgment sentence text. Thus, based on the detection result of the second matching degree, the first evaluation result of the judgment document to be evaluated is generated. If the two are consistent, it means that there is no deviation between the actual judgment sentence and the predicted judgment sentence. If the two are inconsistent, it means that there is a deviation between the actual judgment sentence and the predicted judgment sentence, and the judgment document to be evaluated has quality problems.

[0060] In this embodiment, by obtaining the judgment document to be evaluated, and inputting the judgment document to be evaluated into the machine learning model to extract the case features of the judgment document to be evaluated, and outputting the predicted judgment crime text and predicted judgment type text corresponding to the judgment document to be evaluated, the predicted judgment crime text and predicted judgment type text are detected to match the actual judgment crime text and actual judgment type text in the judgment document to be evaluated, respectively, to determine whether there is a deviation in conviction or type of punishment. The review of the judgment crime text and the judgment type text in the judgment document to be evaluated is realized, which can not only assist the trial management department of the court to accurately and quickly determine the quality of the case, but also improve the efficiency of the trial management department of the court in reviewing criminal cases.

[0061] In one embodiment, the method further includes: when the first matching degree is an incomplete match, generating a determination result that the judgment document to be evaluated is a conviction deviation document.

[0062] Specifically, the judgment result segment of the judgment document to be evaluated includes the text of the actual judgment charge, and the machine learning model is used to output the predicted judgment charge text corresponding to the judgment document to be evaluated. The actual judgment charge text is compared with the predicted judgment charge text, and the first matching degree between the actual judgment charge text and the predicted judgment charge text is detected to determine whether the two are consistent. If the two are inconsistent, it means that there is a conviction deviation between the actual judgment charge and the predicted judgment charge, that is, the judgment document to be evaluated has a quality problem, and a determination result is generated that the judgment document to be evaluated is a conviction deviation document. It can be understood that in the process of reviewing the conviction deviation of a judgment document to be evaluated, it has been determined that the judgment document to be evaluated has a quality problem of conviction deviation, and the subsequent sentence type also has problems. After determining that the judgment document to be evaluated is a conviction deviation document, the review of the judgment document to be evaluated can be stopped.

[0063] In one embodiment, the judgment document to be evaluated also includes the actual sentence value, and the predicted judgment result also includes the predicted sentence value. Among them, the sentence refers to the period during which the deprivation of freedom sentence declared by the court judgment should be executed. The judgment result segment of the judgment document includes the text of the crime, the text of the type of sentence, and the value of the sentence. Then the training process of the machine learning model may include: the computer device obtains a sample set of judgment documents of criminal cases, annotates the judgment documents of criminal cases with the crime, the type of sentence, and the sentence to obtain the labels used for training, and trains the machine learning model using the sample set and the labels to obtain a machine learning model for predicting the judgment result. Specifically, the judgment document to be evaluated is input into the machine learning model, and the machine learning model is used to analyze the case characteristics corresponding to the judgment document to be evaluated, and predictions are made based on the case characteristics obtained by the analysis. The machine learning model outputs the predicted judgment crime text, the predicted type of sentence text, and the predicted sentence value corresponding to the judgment document to be evaluated.

[0064] like Figure 3 As shown, according to the detection result of the second matching degree, a first evaluation result of the judgment document to be evaluated is generated, including:

[0065] S310. When the second matching degree is an incomplete match, a determination result is generated that the judgment document to be evaluated is a document of deviation from the criminal type.

[0066] Specifically, the judgment result segment of the judgment document to be evaluated includes the text of the actual judgment type of punishment, and the machine learning model is used to output the predicted judgment type of punishment text corresponding to the judgment document to be evaluated. The actual judgment type of punishment text is compared with the predicted judgment type of punishment text, and the second matching degree between the actual judgment type of punishment text and the predicted judgment type of punishment text is detected to determine whether the two are consistent. If the two are inconsistent, it means that there is a conviction deviation between the actual judgment type of punishment and the predicted judgment type of punishment, that is, the judgment document to be evaluated has quality problems, and it is determined that the judgment document to be evaluated is a punishment type deviation document. It can be understood that in the process of reviewing the conviction deviation of a judgment document to be evaluated, if it has been determined that the judgment document to be evaluated does not have a conviction deviation, then the judgment document to be evaluated will continue to be reviewed for punishment type deviation. If it has been determined that the judgment document to be evaluated has a punishment type deviation, and the subsequent sentence value is also problematic, then after determining that the judgment document to be evaluated is a punishment type deviation document, the review of the judgment document to be evaluated can be stopped.

[0067] S320, when the second matching degree is a complete match, detecting a sentencing difference value between the actual sentence value and the predicted sentence value;

[0068] S330. Generate a second evaluation result of the judgment document to be evaluated based on the detection result of the sentencing difference value.

[0069] Among them, since the judgment result segment includes not only the text of the actual judgment crime and the text of the actual judgment type, but also the actual judgment sentence value, therefore, in order to ensure the accuracy of the review results and comprehensively review the quality of criminal cases, if there is no conviction deviation and sentence type deviation in a judgment document to be evaluated, it is necessary to further review the sentence value of the judgment document to be evaluated. Specifically, if there is no conviction deviation between the actual judgment crime and the predicted judgment crime, and there is no sentence type deviation between the actual judgment type and the predicted judgment type, the actual sentence value in the judgment document to be evaluated is compared with the predicted sentence value output by the machine learning model, and the sentencing difference value between the actual sentence value and the predicted sentence value is detected to determine whether there is a sentence deviation between the actual sentence and the predicted sentence. Therefore, based on the detection result of the sentencing difference value, the second evaluation result of the judgment document to be evaluated is generated, and the second evaluation result is used to evaluate whether there are quality problems in the judgment document to be evaluated.

[0070] In this embodiment, the predicted judgment crime text, predicted judgment type text, and predicted judgment sentence value are compared with the actual judgment crime text, actual judgment type text, and actual sentence value in the judgment document to be evaluated to determine whether there is a deviation in conviction, type of punishment, or sentence. The machine learning model is used to conduct a comprehensive review of the quality of the judgment document to be evaluated, to identify criminal cases with quality problems, and to ensure the accuracy of the review results. It can not only assist the court's trial management department to accurately and quickly determine the quality of the case, but also improve the efficiency of the court's trial management department in reviewing criminal cases.

[0071] In one embodiment, Figure 4 As shown, according to the detection result of the sentencing difference value, the second evaluation result of the judgment document to be evaluated is generated, including:

[0072] S410. If there is no sentencing difference between the actual sentence value and the predicted sentence value, a determination result is generated that the judgment document to be evaluated is a non-deviation document.

[0073] Among them, a non-deviation document refers to a judgment document in which the actual judgment crime text, actual judgment type text, and actual judgment sentence value in the judgment result section do not deviate from the predicted judgment crime text, predicted judgment type text, and predicted sentence value output by the machine learning model, that is, a non-deviation document is a judgment document without quality problems. Specifically, the actual sentence value is compared with the predicted sentence value to determine whether there is a sentencing difference between the two. If there is no sentencing difference, it indicates that the judgment document to be evaluated has no problems in terms of conviction, type of punishment, and sentence, that is, there is no quality problem, and the judgment document to be evaluated is generated as a judgment result of a non-deviation document.

[0074] S420. If there is a sentencing difference between the actual sentence and the predicted sentence, a third evaluation result of the judgment document to be evaluated is generated based on the sentencing difference.

[0075] Among them, due to practical considerations, such as the trial experience of the trial judge and other factors, relatively small deviations are allowed in the sentence. In order to ensure the credibility of the results of the quality review of criminal cases, when there is a sentencing difference value, it is necessary to further analyze the sentencing difference value to determine whether there is a quality problem of sentence deviation, and since the sentence is a time-related value, the actual sentence value can be compared with the predicted sentence value to obtain the difference between the two, that is, the sentencing difference value of sentence deviation. Specifically, the actual sentence value is compared with the predicted sentence value to determine whether there is a sentence deviation between the two. If there is a sentence deviation, the sentencing difference value between the actual sentence value and the predicted sentence value is calculated, and the sentencing difference value is combined to evaluate whether the existing sentence deviation will lead to a penalty deviation, so as to determine whether there is a quality problem with the judgment document to be evaluated.

[0076] In this embodiment, the judgment document to be evaluated is reviewed based on the difference between the actual sentence value and the predicted sentence value, so as to comprehensively determine whether there are quality problems in the judgment document to be evaluated, thereby improving the accuracy and credibility of the review results.

[0077] In one embodiment, a safety threshold is set between the actual sentence value and the predicted sentence value. According to the detection result of the sentencing difference value, a second evaluation result of the judgment document to be evaluated is generated, including: if the sentencing difference value is less than or equal to the safety threshold, a determination result is generated that the judgment document to be evaluated is a non-deviation document; if the sentencing difference value is greater than the safety threshold, a determination result is generated that the judgment document to be evaluated is a sentence deviation document.

[0078] Among them, as mentioned above, in order to ensure the credibility of the quality evaluation results of criminal cases, deviations are allowed between the predicted sentence value and the actual sentence value, so a safety threshold is set between the actual sentence value and the predicted sentence value. The safety threshold represents the safe range of the difference between the predicted sentence value and the actual sentence value. If the safety threshold is too large, criminal cases with quality problems will be omitted. If the safety threshold is too small, the quality evaluation of criminal cases will be too harsh. The setting of the safety threshold is set according to the actual case situation, and this application does not limit the setting of the safety threshold. Specifically, the sentencing difference value between the actual sentence value and the predicted sentence value is compared with the allowed safety threshold. If the sentencing difference value is less than or equal to the safety threshold, it means that there is no quality problem in the judgment result of the judgment document to be evaluated, and the judgment document to be evaluated is judged to be a non-deviation document. If the sentencing difference value is greater than the safety threshold, it means that there is a quality problem in the judgment result of the judgment document to be evaluated, and the judgment document to be evaluated is judged to be a sentence deviation document.

[0079] Furthermore, the setting of the safety threshold can be set from two perspectives: relative value and absolute value. The safety threshold can include a first safety threshold and a second safety threshold. If the first safety threshold is expressed as an absolute value, the sentencing difference value is expressed as an absolute difference value. If the second safety threshold is expressed as a relative value, the sentencing difference value is expressed as a relative difference value.

[0080] Taking the absolute value as an example, the "safety threshold" is set to 3 months. Assuming that the sentence value of the judgment document to be evaluated is 6 months, and the sentence value predicted by machine learning is 8 months, then the sentencing difference value of the document review is |6-8|=2, and the difference value 2<3, the judgment document to be evaluated is judged to be a non-deviation document. Assuming that the sentence value of the judgment document to be evaluated is 6 months, and the sentence value predicted by machine learning is 5 months, then the sentencing difference value of the document review is |6-5|=1, and the difference value 1<3, the judgment document to be evaluated is judged to be a non-deviation document.

[0081] Taking the relative value as an example, the "safety threshold" is set to 20%. Assuming that the sentence value predicted by machine learning is 10 months, and the sentence value of the judgment document to be evaluated is 9 months, then the sentencing difference value of the document review is |9-10| / 10=10%, and the difference value 10%<20% (safety threshold), and the judgment document to be evaluated is judged to be a non-deviation document. Assuming that the sentence value of the judgment document to be evaluated is 6 months, and the sentence value predicted by machine learning is 7 months, then the sentencing difference value of the document review is |6-5| / 7=14%, and the difference value 14%<20%, and the judgment document to be evaluated is judged to be a non-deviation document.

[0082] In one embodiment, the safety threshold includes a first safety threshold, and the sentencing difference value includes an absolute difference value. If the sentencing difference value is greater than the safety threshold, a determination result is generated that the judgment document to be evaluated is a sentence deviation document, including: if the absolute difference value is greater than the first safety threshold, and the actual sentence value is greater than the predicted sentence value, a determination result is generated that the judgment document to be evaluated is a sentence deviation document with excessively heavy sentences; if the absolute difference value is greater than the first safety threshold, and the actual sentence value is less than the predicted sentence value, a determination result is generated that the judgment document to be evaluated is a sentence deviation document with too light sentences.

[0083] For example, the "first safety threshold" is set to 3 months. Assuming that the sentence value of the judgment document to be evaluated is 4 months, and the sentence value predicted by machine learning is 8 months, then the sentencing difference value of the document review is |4-8|=4, and the difference value 4>3, and the judgment document to be evaluated is judged to be a document with too light a sentence. Assuming that the sentence value of the judgment document to be evaluated is 9 months, and the sentence value predicted by machine learning is 5 months, then the sentencing difference value of the document review is |9-5|=4, and the difference value 4>3, and the judgment document to be evaluated is judged to be a document with too heavy a sentence.

[0084] In one embodiment, the safety threshold includes a second safety threshold, and the sentencing difference value includes a relative difference value. If the sentencing difference value is greater than the safety threshold, a determination result is generated that the judgment document to be evaluated is a sentence deviation document, including: if the relative difference value is greater than the second safety threshold, and the actual sentence value is greater than the predicted sentence value, a determination result is generated that the judgment document to be evaluated is a sentence deviation document with excessively heavy sentences; if the relative difference value is greater than the second safety threshold, and the actual sentence value is less than the predicted sentence value, a determination result is generated that the judgment document to be evaluated is a sentence deviation document with too light sentences.

[0085] For example, the "second safety threshold" is set to 20%. Assuming that the sentence value predicted by machine learning is 10 months, and the sentence value of the judgment document to be evaluated is 6 months, then the sentencing difference value of the document review is |6-10| / 10=40%, and the difference value of 40% is greater than 20% (the second safety threshold), and the judgment document to be evaluated is judged to be a document with a light sentence. Assuming that the sentence value of the judgment document to be evaluated is 8 months, and the sentence value predicted by machine learning is 6 months, then the sentencing difference value of the document review is |8-6| / 6=33%, and the difference value of 33% is greater than 20%, and the judgment document to be evaluated is judged to be a document with a heavy sentence.

[0086] It should be noted that although the embodiment of the present application sets the safety threshold from the perspectives of relative value and absolute value, in the actual review process, the absolute difference value and the relative difference value are selected according to the actual situation. For example, in the review of some serious criminal cases, the setting of the safety threshold can be selected in the form of relative value.

[0087] In one embodiment, the method also includes: saving the review process of non-deviation documents, conviction deviation documents, punishment type deviation documents, and sentence term deviation documents to a case review database in the form of a graphical data structure; and establishing an index based on the trial judge, conviction charge, punishment type, and sentence term.

[0088] Among them, the graphic data structure mainly studies the relationship between shapes and graphic data elements. Specifically, the predicted crime, predicted sentence, and predicted sentence are compared with the actual crime, actual sentence, and actual sentence in the judgment document to be evaluated to determine whether there is a deviation in conviction, type of sentence, or sentence, including:

[0089] The actual conviction charges are compared with the predicted conviction charges. If there is a conviction deviation between the actual conviction charges and the predicted conviction charges, the judgment document to be evaluated is determined to be a conviction deviation document; the conviction deviation document and the corresponding review process are stored in the case review database in the form of a graphical data structure.

[0090] If there is no deviation in conviction, the actual sentence will be compared with the predicted sentence. If there is a deviation between the actual sentence and the predicted sentence, the judgment document to be evaluated is determined to be a sentence deviation document; the sentence deviation document and the corresponding review process are stored in the case review database in the form of a graphical data structure.

[0091] If there is no deviation in the type of sentence, the actual sentence will be compared with the predicted sentence. If there is no deviation between the actual sentence and the predicted sentence, the judgment document to be evaluated is determined to be a non-deviation document; the non-deviation document and the corresponding review process are stored in the case review database in the form of a graphical data structure.

[0092] If there is a deviation in the sentence, the sentencing difference value will be compared with the safety threshold. If the sentencing difference value is less than or equal to the safety threshold, the judgment document to be evaluated is determined to be a non-deviation document; the non-deviation document and the corresponding review process are stored in the case review database in the form of a graphical data structure.

[0093] If there is a deviation in sentence, the sentencing difference value will be compared with the safety threshold. If the sentencing difference value is greater than the safety threshold, the judgment document to be evaluated is determined to be a sentence deviation document; the sentence deviation document and the corresponding review process are stored in the case review database in the form of a graphical data structure.

[0094] Finally, a case review database index is established based on the trial judge, crime, type of sentence and sentence, so as to realize rapid query of case review results.

[0095] In this embodiment, the quality of criminal cases is analyzed intelligently based on a large amount of case data, and the hierarchical relationship of the review process is presented in a graphical structure, which reduces the manual judgment of the trial management department and improves the efficiency and quality of case review.

[0096] In one embodiment, Figure 5 As shown, obtain the judgment documents to be evaluated, including:

[0097] S510, obtaining the case delineation conditions of the judgment document to be evaluated;

[0098] S520. Search in the criminal case database according to the case delineation criteria to obtain the judgment document to be evaluated.

[0099] Among them, the case delineation conditions include at least one of the geographical scope, crime, trial court, and judgment date. Specifically, the user sets at least one of the geographical scope, crime, trial court, and judgment date corresponding to the judgment document to be evaluated through the display interface of the terminal, that is, sets the case delineation conditions. The terminal user triggers the corresponding button, and the terminal initiates an evaluation instruction for the judgment document to be evaluated to the server, and the evaluation instruction carries the case delineation conditions of the judgment document to be evaluated. The server receives the evaluation instruction of the judgment document to be evaluated through the terminal, so that the server obtains the case delineation conditions of the judgment document to be evaluated. The server can pre-store a criminal case library locally, and search in the criminal case library according to the acquired case delineation conditions to obtain the judgment document to be evaluated corresponding to the case delineation conditions.

[0100] In one embodiment, Figure 6 As shown, the present application provides a method for processing judicial documents, which also includes the following steps:

[0101] S602. Obtain case delineation conditions for the judgment document to be evaluated.

[0102] Among them, the case delineation conditions include at least one of the geographical scope, crime, trial court, and judgment date.

[0103] S604. Search in the criminal case database according to the case delineation conditions to obtain the judgment document to be evaluated.

[0104] Among them, the judgment documents to be evaluated include the text of the actual crime, the text of the actual sentence and the actual sentence value.

[0105] S606: Set a safety threshold between the actual sentence value and the predicted sentence value.

[0106] S608. Input the judgment document to be evaluated into the machine learning model for analysis to obtain the predicted judgment result corresponding to the judgment document to be evaluated.

[0107] Among them, the predicted judgment result includes the predicted judgment crime text, the predicted judgment sentence type text and the predicted judgment sentence value.

[0108] S610: Detect the first degree of match between the actual judgment crime text and the predicted judgment crime text.

[0109] S612. When the first matching degree is an incomplete match, a determination result is generated that the judgment document to be evaluated is a conviction deviation document, and the corresponding review process is saved in the case review database in the form of a graphical data structure.

[0110] S614. When the first matching degree is a complete match, detect the second matching degree between the actual sentence text and the predicted sentence text.

[0111] S616. When the second matching degree is an incomplete match, a determination result is generated that the judgment document to be evaluated is a criminal deviation document, and the corresponding review process is saved in the case review database in the form of a graphical data structure.

[0112] S618. When the second matching degree is a complete match, detect the sentencing difference value between the actual sentence value and the predicted sentence value.

[0113] Among them, a safety threshold is set between the actual sentence value and the predicted sentence value.

[0114] S620. If the sentencing difference value is less than or equal to the safety threshold, a determination result is generated that the judgment document to be evaluated is a non-deviation document.

[0115] S622. If the sentencing difference value is greater than the safety threshold, a determination result is generated that the judgment document to be evaluated is a sentence deviation document, and the corresponding review process is saved in the case review database in the form of a graphical data structure.

[0116] Among them, if the safety threshold includes the first safety threshold, the sentencing difference value includes the absolute difference value. Then, if the sentencing difference value is greater than the safety threshold, a determination result is generated that the judgment document to be evaluated is a document with a deviation from the sentence, including: if the absolute difference value is greater than the first safety threshold, and the actual sentence value is greater than the predicted sentence value, then a determination result is generated that the judgment document to be evaluated is a document with a deviation from the sentence that is too heavy; if the absolute difference value is greater than the first safety threshold, and the actual sentence value is less than the predicted sentence value, then a determination result is generated that the judgment document to be evaluated is a document with a deviation from the sentence that is too light.

[0117] If the safety threshold includes a second safety threshold, the sentencing difference value includes a relative difference value. Then, if the sentencing difference value is greater than the safety threshold, a determination result is generated that the judgment document to be evaluated is a document with a deviation from the sentence, including: if the relative difference value is greater than the second safety threshold, and the actual sentence value is greater than the predicted sentence value, then a determination result is generated that the judgment document to be evaluated is a document with a deviation from the sentence that is too heavy; if the relative difference value is greater than the second safety threshold, and the actual sentence value is less than the predicted sentence value, then a determination result is generated that the judgment document to be evaluated is a document with a deviation from the sentence that is too light.

[0118] S624. Establish an index for the review database based on the trial judge, crime, type of sentence and sentence.

[0119] It should be understood that, although the various steps in the above-mentioned flow chart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the above-mentioned flow chart may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0120] In one embodiment, Figure 7 As shown, the present application provides a device for processing a judgment document, and the processing device 700 includes:

[0121] The judgment document acquisition module 710 is used to acquire the judgment document to be evaluated, which includes the text of the actual crime and the text of the actual sentence;

[0122] The judgment prediction module 720 is used to input the judgment document to be evaluated into the machine learning model for analysis to obtain the predicted judgment result corresponding to the judgment document to be evaluated, and the predicted judgment result includes the predicted judgment crime text and the predicted judgment type text;

[0123] A first detection module 730 is used to detect a first matching degree between the actual judgment crime text and the predicted judgment crime text;

[0124] The second detection module 740 is used to detect the second matching degree between the actual sentence type text and the predicted sentence type text when the first matching degree is a complete match;

[0125] The evaluation result generation module 750 is used to generate a first evaluation result of the judgment document to be evaluated based on the detection result of the second matching degree.

[0126] For the specific definition of the processing device of the judgment document, please refer to the definition of the processing method of the judgment document above, which will not be repeated here. The various modules in the above-mentioned processing device of the judgment document can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules. For example: the above-mentioned processing device 700 includes a processor and a memory, and the above-mentioned judgment document acquisition module 710, judgment prediction module 720, first detection module 730, second detection module 740, evaluation result generation module 750, etc. are all stored in the memory as program modules, and the processor executes the above-mentioned program modules stored in the memory to realize the corresponding functions.

[0127] The processor includes a kernel, which retrieves the corresponding program module from the memory. One or more kernels can be set, and the judgment document to be evaluated can be input into the machine learning model for analysis by adjusting the kernel parameters, and the judgment crime text, judgment type text and judgment sentence value corresponding to the judgment document to be evaluated can be predicted; the predicted judgment crime text, judgment type text and judgment sentence value can be compared with the actual judgment crime text, judgment type text and judgment sentence value, and criminal cases with quality problems can be found, which can assist the court's trial management department to accurately and quickly determine the case quality and improve the efficiency of the court's trial management department in reviewing criminal cases.

[0128] An embodiment of the present application provides a storage medium on which a program is stored, and when the program is executed by a processor, the method for processing the judicial document is implemented.

[0129] An embodiment of the present application provides a processor, which is used to run a program, wherein the method for processing the judicial document is executed when the program is running.

[0130] The embodiment of the present application provides a computer device, which includes at least one processor, at least one memory connected to the processor, and a bus; wherein the processor and the memory communicate with each other through the bus; the processor is used to call program instructions in the memory to execute the above-mentioned method for processing judgment documents. The device in this article can be a server, PC, PAD, mobile phone, etc.

[0131] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program for initializing the following method steps:

[0132] Obtaining the judgment document to be evaluated, wherein the judgment document to be evaluated includes the text of the actual crime and the text of the actual sentence;

[0133] Input the judgment document to be evaluated into the machine learning model for analysis to obtain the predicted judgment result corresponding to the judgment document to be evaluated, wherein the predicted judgment result includes the predicted judgment crime text and the predicted judgment type text;

[0134] Detecting a first degree of match between the actual judgment crime text and the predicted judgment crime text;

[0135] When the first matching degree is a complete match, detecting a second matching degree between the actual sentence text and the predicted sentence text;

[0136] Based on the detection result of the second matching degree, a first evaluation result of the judgment document to be evaluated is generated.

[0137] In one embodiment, the method further includes: when the first matching degree is an incomplete match, generating a determination result that the judgment document to be evaluated is a conviction deviation document.

[0138] In one embodiment, the judgment document to be evaluated further includes an actual sentence value, and the predicted judgment result further includes a predicted sentence value; and generating a first evaluation result of the judgment document to be evaluated based on the detection result of the second matching degree includes:

[0139] When the second matching degree is an incomplete match, a determination result is generated that the judgment document to be evaluated is a document of deviation from the criminal type;

[0140] When the second matching degree is a complete match, detecting a sentencing difference value between the actual sentence value and the predicted sentence value;

[0141] Based on the detection result of the sentencing difference value, a second evaluation result of the judgment document to be evaluated is generated.

[0142] In one embodiment, a safety threshold is set between the actual sentence value and the predicted sentence value; and generating a second evaluation result of the judgment document to be evaluated based on the detection result of the sentencing difference value includes:

[0143] If the sentencing difference value is less than or equal to the safety threshold, a determination result is generated that the judgment document to be evaluated belongs to the non-deviation document;

[0144] If the sentencing difference value is greater than the safety threshold, a determination result is generated that the judgment document to be evaluated belongs to the sentence deviation document.

[0145] In one embodiment, the safety threshold includes a first safety threshold, and the sentencing difference value includes an absolute difference value; if the sentencing difference value is greater than the safety threshold, generating a determination result that the judgment document to be evaluated belongs to the sentence deviation document, including:

[0146] If the absolute difference value is greater than the first safety threshold, and the actual sentence value is greater than the predicted sentence value, a determination result is generated that the judgment document to be evaluated is a document with excessively severe sentence deviation;

[0147] If the absolute difference value is greater than the first safety threshold, and the actual sentence value is less than the predicted sentence value, then a determination result is generated that the judgment document to be evaluated is a deviation from the document with too light a sentence.

[0148] In one embodiment, the safety threshold includes a second safety threshold, and the sentencing difference value includes a relative difference value; if the sentencing difference value is greater than the safety threshold, generating a determination result that the judgment document to be evaluated belongs to the sentence deviation document, includes:

[0149] If the relative difference value is greater than the second safety threshold, and the actual sentence value is greater than the predicted sentence value, a determination result is generated that the judgment document to be evaluated is a document with excessively severe sentence deviation;

[0150] If the relative difference value is greater than the second safety threshold, and the actual sentence value is less than the predicted sentence value, then a determination result is generated that the judgment document to be evaluated is a deviation document with too light a sentence.

[0151] In one embodiment, the method further includes: saving the review process of the non-deviation document, the conviction deviation document, the punishment type deviation document, and the sentence deviation document to a case review database in the form of a graphical data structure; and establishing an index based on the trial judge, the crime sentenced, the punishment type sentenced, and the sentence sentenced.

[0152] In one embodiment, obtaining the judgment document to be evaluated includes: obtaining case delineation conditions of the judgment document to be evaluated, the case delineation conditions including at least one of geographical scope, crime, trial court, and judgment date; searching in the criminal case database according to the case delineation conditions to obtain the judgment document to be evaluated.

[0153] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0154] In a typical configuration, such as Figure 8 As shown, the device includes one or more processors (CPU), memory and bus. The device may also include input / output interface, network interface and the like.

[0155] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip. The memory is an example of a computer-readable medium.

[0156] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0157] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0158] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0159] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for processing judicial documents, characterized in that: The method comprises: Obtaining the judgment document to be evaluated, wherein the judgment document to be evaluated includes the text of the actual crime, the text of the actual sentence, and the actual sentence value; Input the judgment document to be evaluated into the machine learning model for analysis to obtain the predicted judgment result corresponding to the judgment document to be evaluated, the predicted judgment result includes the predicted judgment crime text, the predicted judgment type text and the predicted judgment sentence value, and a safety threshold is set between the actual judgment sentence value and the predicted judgment sentence value; the safety threshold includes a first safety threshold and a second safety threshold; Detecting a first degree of match between the actual judgment crime text and the predicted judgment crime text; When the first matching degree is a complete match, detecting a second matching degree between the actual sentence text and the predicted sentence text; When the first matching degree is an incomplete match, a determination result is generated that the judgment document to be evaluated is a conviction deviation document; When the second matching degree is an incomplete match, a determination result is generated that the judgment document to be evaluated is a document of deviation from the criminal type; When the second matching degree is a complete match, detecting a sentencing difference value between the actual sentence value and the predicted sentence value; the sentencing difference value includes an absolute difference value and a relative difference value; If the sentencing difference value is less than or equal to the safety threshold, a determination result is generated that the judgment document to be evaluated is a non-deviation document; If the sentencing difference value is greater than the safety threshold, a determination result is generated that the judgment document to be evaluated is a document that deviates from the sentence, including: if the absolute difference value is greater than the first safety threshold, and the actual sentence value is greater than the predicted sentence value, a determination result is generated that the judgment document to be evaluated is a document that deviates from the sentence by excessive severity; if the absolute difference value is greater than the first safety threshold, and the actual sentence value is less than the predicted sentence value, a determination result is generated that the judgment document to be evaluated is a document that deviates from the sentence by excessive severity; if the relative difference value is greater than the second safety threshold, and the actual sentence value is greater than the predicted sentence value, a determination result is generated that the judgment document to be evaluated is a document that deviates from the sentence by excessive severity; if the relative difference value is greater than the second safety threshold, and the actual sentence value is less than the predicted sentence value, a determination result is generated that the judgment document to be evaluated is a document that deviates from the sentence by excessive severity.

2. The method according to claim 1, characterized in that: The method further comprises: The review process of the non-deviation document, the conviction deviation document, the punishment type deviation document and the sentence deviation document is saved in the case review database in the form of a graphic data structure; and an index is established based on the trial judge, the crime sentenced, the punishment type sentenced and the sentence sentenced.

3. The method according to claim 1, characterized in that The obtaining of the judgment document to be evaluated includes: Obtain case delineation conditions for the judgment documents to be evaluated; The judgment document to be evaluated is obtained by searching in the criminal case database according to the case delineation conditions.

4. The method according to claim 1, characterized in that: The method further comprises: The predicted crime text, the predicted type of punishment text and the predicted sentence value are respectively compared with the actual crime text, the actual type of punishment text and the actual sentence value in the judgment document to be evaluated to determine whether there is a deviation in conviction, deviation in type of punishment and deviation in sentence in the judgment document to be evaluated.

5. The method according to claim 4, characterized in that The step of comparing the predicted crime text, the predicted sentence type text and the predicted sentence value with the actual crime text, the actual sentence type text and the actual sentence value in the judgment document to be evaluated, respectively, to determine whether the judgment document to be evaluated has a deviation in conviction, a deviation in sentence type and a deviation in sentence, includes: Compare the actual judgment crime text with the predicted judgment crime text, and if there is a conviction deviation between the actual judgment crime text and the predicted judgment crime text, determine that the judgment document to be evaluated is a conviction deviation document; If there is no conviction deviation, the actual sentence text is compared with the predicted sentence text. If there is a deviation between the actual sentence text and the predicted sentence text, the judgment document to be evaluated is determined to be a sentence deviation document; If there is no deviation in the type of sentence, the actual sentence value is compared with the predicted sentence value. If there is no deviation in the sentence between the actual sentence value and the predicted sentence value, it is determined that the judgment document to be evaluated is a document without deviation; If there is a deviation in the sentence, the sentencing difference value is compared with the safety threshold. If the sentencing difference value is less than or equal to the safety threshold, the judgment document to be evaluated is determined to be a document without deviation; If there is a deviation in the sentence, the sentencing difference value is compared with the safety threshold. If the sentencing difference value is greater than the safety threshold, it is determined that the judgment document to be evaluated is a sentence deviation document.

6. A device for processing judicial documents, characterized in that: The device comprises: A judgment document acquisition module is used to acquire the judgment document to be evaluated, wherein the judgment document to be evaluated includes the text of the actual crime, the text of the actual sentence, and the actual sentence value; A judgment prediction module, used for inputting the judgment document to be evaluated into a machine learning model for analysis, and obtaining a predicted judgment result corresponding to the judgment document to be evaluated, wherein the predicted judgment result includes a predicted judgment crime text, a predicted judgment type text, and a predicted judgment sentence value, and a safety threshold is set between the actual judgment sentence value and the predicted judgment sentence value; the safety threshold includes a first safety threshold and a second safety threshold; A first detection module, used to detect a first matching degree between the actual judgment crime text and the predicted judgment crime text; A second detection module is used to detect a second matching degree between the actual sentence text and the predicted sentence text when the first matching degree is a complete match; A first evaluation result generating module, configured to generate a determination result that the judgment document to be evaluated is a conviction deviation document when the first matching degree is an incomplete match; A second evaluation result generating module, configured to generate a determination result that the judgment document to be evaluated is a criminal deviation document when the second matching degree is an incomplete match; A third evaluation result generating module is used for detecting a sentencing difference value between the actual sentence value and the predicted sentence value when the second matching degree is a complete match; the sentencing difference value includes an absolute difference value and a relative difference value; The third evaluation result generating module is further configured to generate a determination result that the judgment document to be evaluated is a non-deviation document if the sentencing difference value is less than or equal to the safety threshold; The third evaluation result generating module is also used for generating a determination result that the judgment document to be evaluated is a document deviating from the sentence if the sentencing difference value is greater than the safety threshold; the third evaluation result generating module is also used for generating a determination result that the judgment document to be evaluated is a document deviating from the sentence with excessive punishment if the absolute difference value is greater than the first safety threshold and the actual sentence value is greater than the predicted sentence value; if the absolute difference value is greater than the first safety threshold and the actual sentence value is less than the predicted sentence value, then generating a determination result that the judgment document to be evaluated is a document deviating from the sentence with excessive punishment; if the relative difference value is greater than the second safety threshold and the actual sentence value is greater than the predicted sentence value, then generating a determination result that the judgment document to be evaluated is a document deviating from the sentence with excessive punishment; if the relative difference value is greater than the second safety threshold and the actual sentence value is less than the predicted sentence value, then generating a determination result that the judgment document to be evaluated is a document deviating from the sentence with excessive punishment.

7. A computer device, characterized in that: It includes at least one processor, and at least one memory and a bus connected to the processor; wherein the processor and the memory communicate with each other through the bus; the processor is used to call the program instructions in the memory to execute the steps of the method for processing judicial documents described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for processing judicial documents described in any one of claims 1 to 5 are implemented.