WO2021051865A1

WO2021051865A1 - Case recommendation method and device, apparatus, and computer readable storage medium

Info

Publication number: WO2021051865A1
Application number: PCT/CN2020/093108
Authority: WO
Inventors: 邓俊豪; 聂宇昕; 徐冰; 陈晨; 汪伟
Original assignee: 平安科技（深圳）有限公司
Priority date: 2019-09-18
Filing date: 2020-05-29
Publication date: 2021-03-25
Also published as: CN110795566A

Abstract

The present invention relates to the technical field of data analysis in artificial intelligence. Provided are a case recommendation method and device, an apparatus, and a computer readable storage medium. The method comprises: acquiring judgement data of a target case, and using a preset focus of disputes determination model to determine the focus of disputes of the target case according to the judgement data; constructing a first knowledge graph of the target case according to the judgement data and the focus of disputes; determining, from a pre-stored closed case base, and according to the focus of disputes, a candidate case set to be recommended; acquiring second knowledge graphs corresponding to respective candidate cases in the candidate case set, and performing calculation to obtain levels of similarity between the first knowledge graph and the respective second knowledge graphs; and determining similar cases with similar decisions from the candidate case set according to the levels of similarity between the first knowledge graph and the respective second knowledge graphs. The present invention relates to data analysis and knowledge graphs, and can be used to accurately determine similar cases with similar decisions.

Description

Case recommendation method, device, equipment and computer readable storage medium

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on September 18, 2019, the application number is 201910883521.9, and the invention title is "case recommendation method, device, equipment, and computer-readable storage medium". The entire content of the Chinese patent application is approved. The reference is incorporated in this application.

Technical field

This application relates to the technical field of data analysis in artificial intelligence, and in particular to a case recommendation method, device, equipment, and computer-readable storage medium.

Background technique

At present, the focus of the case can usually represent the entire case. Therefore, the judge needs to determine the focus of the dispute based on the plaintiff’s petition, the defendant’s defense, and the evidence provided by both parties. The judgment can be completed by resolving the focus of the dispute, and the facts need to be considered when resolving the focus of the dispute. Factors, evidence, laws and regulations, etc., while the facts, evidence, and laws and regulations are complex, there will be situations where the focus of the dispute is the same but the results of the case are different, which is not convenient for the judge to decide the case.

In order to solve the above-mentioned problems, judges refer to similar cases that have been tried before to effectively regulate and restrict discretion to ensure that similar cases are basically uniform in the application of law, basically the same criteria for judgment, and basically consistent judgment results. However, the inventor realizes that there are many cases that have been tried before, and it takes more time to find similar cases, and the judges subjectively determine the logical relationship between the cases, and the judgments of similar cases may not be found. Applicable to the current case, which affects the efficiency of judgment. Therefore, how to accurately determine the same case in the same case, ensure the same case and improve the efficiency of the case is a problem that needs to be solved urgently.

technical problem

The main purpose of this application is to provide a case recommendation method, device, equipment, and computer-readable storage medium, aiming to accurately determine the case of the same case, ensure the same case and improve the efficiency of the case.

Technical solutions

In order to achieve the foregoing objectives, in the first aspect, this application provides a case recommendation method, which includes the following steps:

Obtain the judgment data of the target case, and determine the dispute focus of the target case based on the judgment data based on a preset dispute focus determination model;

Construct a first knowledge graph of the target case according to the judgment data and the dispute focus;

According to the focus of the dispute, determine the candidate case set to be recommended from the pre-stored judged case database;

Acquiring a second knowledge graph corresponding to each candidate case in the candidate case set, and calculating the similarity between the first knowledge graph and each of the second knowledge graphs;

According to the degree of similarity between the first knowledge graph and each of the second knowledge graphs, co-judge cases are determined from the set of candidate cases.

In the second aspect, this application also provides a case recommendation device, the case recommendation device including:

The dispute focus determination module is used to obtain the judgment data of the target case, and based on the preset dispute focus determination model, determine the dispute focus of the target case according to the judgment data;

The graph construction module is used to construct the first knowledge graph of the target case according to the judgment data and the dispute focus;

The candidate case determination module is used to determine the candidate case set to be recommended from the pre-stored judged case database according to the focus of the dispute;

A calculation module, configured to obtain a second knowledge graph corresponding to each candidate case in the candidate case set, and calculate the similarity between the first knowledge graph and each of the second knowledge graphs;

The case determination module is configured to determine the same case and concurrent judgment cases from the set of candidate cases according to the similarity between the first knowledge graph and each of the second knowledge graphs.

In a third aspect, the present application also provides a computer device that includes a processor, a memory, and a computer program stored on the memory and executable by the processor, wherein the computer program is When the processor executes, a case recommendation method is implemented, where the case recommendation method includes:

In a fourth aspect, this application also provides a computer-readable storage medium with a computer program stored on the computer-readable storage medium, and when the computer program is executed by a processor, a case recommendation method is implemented, wherein: The recommended method for the case includes the following steps:

Beneficial effect

This application provides a case recommendation method, device, equipment, and computer-readable storage medium, which can accurately determine the same case and the same case.

Description of the drawings

FIG. 1 is a schematic flowchart of a case recommendation method provided by an embodiment of the application;

Figure 2 is a schematic diagram of the knowledge graph in an embodiment of the application;

FIG. 3 is a schematic flowchart of another case recommendation method provided by an embodiment of the application;

FIG. 4 is a schematic block diagram of a case recommendation device provided by an embodiment of the application;

Fig. 5 is a schematic block diagram of another case recommendation device provided by an embodiment of the application;

FIG. 6 is a schematic block diagram of the structure of a computer device related to an embodiment of this application.

The best mode of the present invention

In order to solve the foregoing problems, embodiments of the present application provide a case recommendation method, device, computer equipment, and computer-readable storage medium. Among them, the case recommendation method can be applied to a server, and the server can be a single server or a server cluster composed of multiple servers.

Please refer to FIG. 1, which is a schematic flowchart of a case recommendation method provided by an embodiment of the application.

As shown in Fig. 1, the case recommendation method includes steps S101 to S105.

Step S101: Obtain the judgment data of the target case, and determine the dispute focus of the target case based on the judgment data based on a preset dispute focus determination model.

Among them, both parties to the case upload the judgment data of the pending case, defense text, and evidence information to the server through the terminal device, or both parties to the case upload the appeal text and defense text of the pending case offline Judgment data such as evidence and evidence information are submitted to the court, and the court staff upload the judgment data submitted by both parties to the case to the server through the terminal device, and the server stores the judgment data of the outstanding cases.

The judgment data includes the petition text, defense text and evidence information. The petition text includes the plaintiff's information and the petition's opinions. The defense text includes the defendant's information and the defense's opinions. The evidence information includes but is not limited to documentary evidence and material evidence. , Audiovisual materials, witness testimony, statements by the parties and conclusions of the appraisal. Evidence attribute refers to the characteristic attribute of the evidence, such as whether the IOU has the signature of the borrower.

When the triggered co-case recommendation instruction is monitored, according to the co-case recommendation instruction, the unconfirmed cases that need to be recommended are determined, and the unconfirmed cases that need to be recommended as the target cases, and then the judgment data of the target cases are obtained. Among them, the trigger mode of the co-case recommendation instruction includes real-time trigger and timing trigger. The real-time trigger is when the co-case recommendation request sent by the terminal device is monitored, the unconfirmed cases that need to be recommended are obtained from the co-case recommendation request of the case. The case number of, triggers the recommendation instruction of the same case that contains the case number; timing trigger sets a certain time task for the server, and obtains a case number from the recommendation queue of the same case through the timed task, and triggers the case reasoning instruction containing the case number .

Specifically, the server extracts the plaintiff request information, the defendant's defense information, and the evidence information from the judgment data; based on the preset dispute focus determination model, the preset number is determined based on the plaintiff's request information, the defendant's defense information, and the evidence information. The candidate dispute focus and the output probability value of each candidate dispute focus; the candidate dispute focus with the largest output probability value is regarded as the dispute focus of the target case. Wherein, the output probability value is the probability value that the dispute focus output of the dispute focus determination model is the candidate dispute focus, and the server stores the first regular expression for extracting the original claim information and the second for extracting the defendant’s defense information. Regular expressions, the first regular expression can be used to extract the original claim information from the judgment data, and the second regular expression can be used to extract the defendant’s defense information from the judgment data.

Among them, the method of extracting evidence information is specifically: extracting evidence sentences and the evidence items in each evidence sentence from the judgment data through the evidence item extraction model; determining the evidence category of each evidence item and the evidence of each evidence category Attributes, and use each evidence item, the evidence category of each evidence item, and the evidence attributes of each evidence category as evidence information.

In one embodiment, the method for determining the evidence category is specifically: obtaining the evidence category of the evidence item and each evidence sub-category under the evidence category from a preset evidence classification table, and calculating the evidence item by a similarity formula The similarity between the evidence keywords corresponding to each evidence sub-category under the evidence category, and then the evidence sub-category with the largest similarity is determined as the evidence category of the evidence item. It should be noted that the evidence classification table can be set based on actual conditions, which is not specifically limited in this application.

In one embodiment, the method for determining the attributes of the evidence is specifically: for each type of evidence, the judgment data is traversed to determine the context information of each type of evidence; and the mapping relationship between the pre-stored evidence attributes and the evidence keywords is searched. , Obtain the target evidence sentence containing the evidence keyword from the context information of each type of evidence, and obtain the evidence attribute group corresponding to the target evidence sentence; after concatenating the target evidence sentence and the evidence attribute group, input the similarity calculation The model calculates the similarity between each evidence attribute in the evidence attribute group and the target evidence sentence, and uses the evidence attribute with the highest similarity as the evidence attribute of the corresponding evidence category to obtain the evidence attribute of each evidence category. It should be noted that the above-mentioned mapping relationship table between the evidence attributes and the evidence keywords can be set based on actual conditions, which is not specifically limited in this application.

Among them, the above-mentioned evidence item extraction model is through AutoNER (Auto Named Entity Recognition (Auto Named Entity Recognition) is obtained by training based on manually labeled sample data. The AutoNER model uses the word vector trained for the ruling document as an embedding. Because the accuracy of the automatic remote labeling data is too low, this module is abandoned and used instead. For training on labeled data, the accuracy of manually labeled sample data is high. In addition, in order to prevent over-fitting, the method of data enhancement is used during training, that is, random replacement of no more than 3 words and/or in the evidence sentence Reverse the order of the words in the evidence sentence. The above-mentioned similarity calculation model is obtained by retraining the pre-training model of the BERT model based on legal corpus, reducing the encoder module (encoding module) of the BERT model to 3 layers, and adjusting the sentence length to achieve time optimization. In addition, for the training task, the interference sample data is obtained, and the model is trained based on the interference sample data, and the classification layer is connected to the position of the encoder layer. The AutoNER model is a model that can automatically label data and train named entity recognition without manual labeling. The BERT model is the first deep, bidirectional and unsupervised language representation model.

Step S102: Construct a first knowledge graph of the target case according to the judgment data and the dispute focus.

Specifically, the basic case knowledge information and evidence knowledge information are extracted from the judgment data, and the focus of dispute, basic case knowledge information and evidence knowledge information are taken as the case knowledge information of the target case, and then obtained from the knowledge base of preset laws The knowledge of the law that matches the knowledge information of the case, and based on the knowledge information of the case and the knowledge of the law, construct the knowledge map of the target case, which is recorded as the first knowledge map.

Among them, the method of constructing the knowledge graph is specifically as follows: the law in the knowledge of the law and the plaintiff, defendant, petition, argument, dispute focus, fact elements, and evidence in the knowledge information of the case are used as the entity nodes of the knowledge graph, And from the legal knowledge and case knowledge information to obtain the relationship and attributes between each entity node (plaintiff, defendant, claim point of view, argued point of view, dispute focus, factual elements, evidence and specific value of the legal provisions), and then Based on the entity node, the relationship between the entity node and the attribute of the entity node, the knowledge graph of the target case is constructed.

Please refer to Figure 2. Figure 2 is a schematic diagram of the knowledge graph in the embodiment of the application. As shown in Figure 2, the entity nodes of the knowledge graph are the plaintiff, the defendant, the opinion of the appeal, the argued opinion, the focus of the dispute, the factual elements, and the evidence And the law, and the fact elements include small element 1, small element 2, and small element 3. Small element 1 corresponds to law 1, small element 2 corresponds to law 2, and small element 3 corresponds to law 3.

Step S103: According to the dispute focus, a candidate case set to be recommended is determined from the pre-stored sentenced case database.

The server also determines the candidate case set to be recommended from the pre-stored sentenced case database according to the dispute focus of the target case, that is, it traverses the knowledge map of all the sentenced cases in the stored sentenced case database, and obtains that the knowledge map includes the dispute Focus on the sentenced cases, and collect each of the sentenced cases obtained to form a sentenced case set, and use the sentenced case set as a candidate case set to be recommended.

Among them, the server stores the knowledge map of the judged case, which is recorded as the second knowledge map. The case knowledge map stores case knowledge information and legal knowledge information. The case knowledge information includes basic case knowledge, dispute focus knowledge, and fact elements Knowledge and evidence knowledge. Basic case knowledge includes but is not limited to litigation figures, litigation companies, litigation participant relationships, plaintiffs, defendants, petition and defense opinions, dispute focus knowledge includes dispute focus, and evidence knowledge includes evidence items and types of evidence And evidence attributes. Evidence attribute refers to the characteristic attribute of the evidence, such as whether the IOU has the signature of the borrower. There is at least one fact in every plaintiff request, such as "A requests B to repay the money". The fact is "A lent money to B and B does not repay it."

The following takes a sentenced case as an example to explain how to construct the knowledge graph of a sentenced case. Specifically, obtain the judgment document of the judged case, and extract the case knowledge from the judgment document to obtain the structured case knowledge information and legal knowledge information of the judged case, and then according to the case knowledge information and legal knowledge information, Construct a knowledge graph of convicted cases. Among them, the ruling document includes the relationship between the characters, the origin of the case, the trial process, the facts, and the reason and basis of the judgment. The character relationship part includes the basic information of the parties, the basic information of the entrusted agents ad litem, and the litigation status of the parties; the origin part of the case includes information about the petition and plea, the part of the trial process is the court record, and the fact part includes the plaintiff’s litigation request, Facts and reasons, the facts and reasons of the defendant’s defense, the facts found by the court and the evidence on which the case was determined, the reasons and basis of the judgment include the relationship between the evidence and the focus of the dispute, and the relationship between the basis of the judgment and the legal provisions.

Step S104: Obtain a second knowledge graph corresponding to each candidate case in the candidate case set, and calculate the similarity between the first knowledge graph and each second knowledge graph.

After determining the candidate case set, the server obtains the second knowledge graph corresponding to each candidate case in the candidate case set, and calculates the similarity between the first knowledge graph and each second knowledge graph. It should be noted that the similarity between the first knowledge graph and each second knowledge graph is the similarity between the target case and each candidate case.

In an embodiment, the server obtains the first case knowledge information from the first knowledge graph, and obtains the respective corresponding second case knowledge information from each second knowledge graph; according to the first case knowledge information and each second knowledge graph; Case knowledge information, calculate the similarity between the first knowledge graph and each second knowledge graph. The first case knowledge information includes attribute information and relationship information of each entity node in the first knowledge graph, and the second knowledge information includes attribute information and relationship information of each entity node in the second knowledge graph.

In an embodiment, the server calculates the target similarity between each entity node in the first knowledge graph and the corresponding entity node in each second knowledge graph based on the first case knowledge information and each second case knowledge information ; According to the respective preset coefficients corresponding to each entity node in the first knowledge graph and the similarity of each target, the similarity between the first knowledge graph and each second knowledge graph is calculated.

Take a second knowledge graph as an example to illustrate the calculation of target similarity. Specifically, obtain the attribute information and/or relationship information of each first entity node from the first case knowledge information, and obtain the second case knowledge information To obtain the attribute information and/or relationship information of each second entity node; according to the attribute information of each first entity node and the attribute information of each second entity node, calculate each first entity node and the corresponding second entity node The first degree of similarity between the entity nodes; and/or calculate the relationship between each first entity node and the corresponding second entity node based on the relationship information of each first entity node and the relationship information of each second entity node The second degree of similarity; take the first degree of similarity and/or the second degree of similarity between each first entity node and the corresponding second entity node as each entity node and each second knowledge in the first knowledge graph The target similarity between corresponding entity nodes in the graph.

Among them, the attribute information is the attributes of the entity nodes in the knowledge graph, such as the specific parameters of the entity nodes such as dispute focus, fact elements, and evidence, and the relationship information is the relationship information between the entity nodes in the knowledge graph, such as dispute focus and facts The relationship between the elements, through the attribute information, can get the semantic similarity between the cases, and the logical similarity between the cases can be obtained through the relationship information. The credibility and accuracy of similarity can be improved.

Wherein, when the target similarity is the first similarity and the second similarity, the calculation method of the similarity between the first knowledge graph and a second knowledge graph is specifically: multiplying each first similarity by its corresponding After the first weight coefficient of, is accumulated to obtain the first target similarity between the first knowledge graph and the second knowledge graph; each second similarity is multiplied by the corresponding second weight coefficient and then accumulated to obtain the first The second target similarity between a knowledge graph and the second knowledge graph; calculate the sum of the first target similarity and the second target similarity, and use the sum of the first target similarity and the second target similarity as the first The similarity between the knowledge graph and the second knowledge graph. Through attribute information and relationship information, the semantic and logical similarity between cases can be comprehensively considered, and the credibility and accuracy of similarity can be further improved.

It should be noted that the above-mentioned first weight coefficient and second weight coefficient can be set based on actual conditions, the first weight coefficient of each physical node can be the same or different, and the second weight coefficient of each physical node can be the same , Can also be different, and this application is not specifically limited.

Step S105, according to the similarity between the first knowledge graph and each of the second knowledge graphs, determine the same case and concurrent judgment cases from the set of candidate cases.

Specifically, the similarity between the first knowledge graph and each second knowledge graph is taken as the similarity between the target case and each candidate case in the candidate case set, and in the order of the magnitude of the similarity, the candidate Each candidate case in the case set is sorted, and the candidate case with the highest ranking is regarded as a co-judgment case, and the candidate case corresponding to the highest similarity is regarded as a co-judgment case.

The case recommendation method provided in the above embodiments is based on the dispute focus determination model, and the dispute focus of the target case can be accurately determined according to the judgment data, and the knowledge map of the target case is constructed based on the judgment data and the dispute focus, and according to the dispute focus, Determine the set of candidate cases to be recommended, and then calculate the similarity between the knowledge graph of the target case and the knowledge graph corresponding to each candidate case, and according to the similarity between the knowledge graph of the target case and the knowledge graph corresponding to each candidate case Degree, can accurately determine the same case and the same sentence.

Please refer to FIG. 3, which is a schematic flowchart of another case recommendation method provided by an embodiment of the application. As shown in FIG. 3, the case recommendation method includes steps S201 to 206.

Step S201: Obtain the judgment data of the target case, and determine the dispute focus of the target case based on the judgment data based on a preset dispute focus determination model.

When the triggered co-case recommendation instruction is monitored, according to the co-case recommendation instruction, the unconfirmed cases that need to be recommended are determined, and the unconfirmed cases that need to be recommended as the target cases, and then the judgment data of the target cases are obtained.

Specifically, the server extracts the plaintiff request information, the defendant's defense information, and the evidence information from the judgment data; based on the preset dispute focus determination model, the preset number is determined based on the plaintiff's request information, the defendant's defense information, and the evidence information. The candidate dispute focus and the output probability value of each candidate dispute focus; the candidate dispute focus with the largest output probability value is regarded as the dispute focus of the target case.

Step S202: Construct a first knowledge graph of the target case according to the judgment data and the dispute focus.

Step S203: Calculate the similarity between the dispute focus and the dispute focus of each judged case in the prestored judged case database.

The server calculates the similarity between the dispute focus of the target case and the dispute focus of each sentenced case in the pre-stored sentenced case database, that is, encodes each word in the dispute focus of the target case to obtain the first vector, and then Each word in the dispute focus of each judged case is coded to obtain the corresponding second vector, and then the cosine similarity between the first vector and each second vector is calculated, and the first vector is compared with each second vector. The cosine similarity between the vectors is taken as the similarity between the dispute focus of the target case and the dispute focus of each judged case.

In one embodiment, the server counts the total number of cases of judged cases, and determines the number of concurrent threads according to the total number of cases, that is, obtains the mapping relationship table between the number of pre-stored cases and the number of concurrent threads, and queries the mapping relationship table , Get the number of concurrent threads corresponding to the total number of cases; According to the number of concurrent threads, call the corresponding number of idle threads in the preset thread pool to concurrently calculate the similarity between the dispute focus and the dispute focus of each judged case. It should be noted that the mapping relationship table between the number of cases and the number of concurrent threads and the number of threads in the thread pool can be set based on actual conditions, and this application does not specifically limit this. By concurrently calculating the similarity between the dispute focus and the dispute focus of each judged case through multiple threads, the calculation speed of the similarity can be improved.

Step S204: According to the similarity between the dispute focus and the dispute focus of each judged case, determine the candidate case set to be recommended from the pre-stored judged case database.

After calculating the similarity between the dispute focus of the target case and the dispute focus of each judged case, each judged case whose similarity is greater than the preset similarity threshold is written into the preset candidate case blank set, To form a set of candidate cases to be recommended. It should be noted that the aforementioned similarity threshold may be set based on actual conditions, which is not specifically limited in this application.

Step S205: Obtain a second knowledge graph corresponding to each candidate case in the candidate case set, and calculate the similarity between the first knowledge graph and each second knowledge graph.

Step S206: According to the similarity between the first knowledge graph and each of the second knowledge graphs, determine the same case and the same judgment case from the set of candidate cases.

The case recommendation method provided by the above embodiment can determine the candidate case set to be recommended more accurately based on the similarity between the dispute focus of the target case and the dispute focus of each sentenced case, and is based on the knowledge map of the target case and The similarity between the knowledge graphs of each candidate case can further accurately determine cases similar to the target case, ensure the same case and improve the efficiency of judgment.

Please refer to FIG. 4, which is a schematic block diagram of a case recommendation apparatus provided by an embodiment of the application.

As shown in FIG. 4, the case recommendation device 300 includes: a dispute focus determination module 301, a graph construction module 302, a candidate case determination module 303, a calculation module 304, and a case determination module 305.

The dispute focus determination module 301 is used to obtain the judgment data of the target case, and based on a preset dispute focus determination model, determine the dispute focus of the target case according to the judgment data;

The graph construction module 302 is configured to construct a first knowledge graph of the target case according to the judgment data and the dispute focus;

The candidate case determination module 303 is configured to determine the candidate case set to be recommended from the pre-stored judged case database according to the focus of the dispute;

The calculation module 304 is configured to obtain a second knowledge graph corresponding to each candidate case in the candidate case set, and calculate the similarity between the first knowledge graph and each second knowledge graph;

The case determination module 305 is configured to determine the same-sentence cases from the set of candidate cases according to the similarity between the first knowledge graph and each of the second knowledge graphs.

In one embodiment, the dispute focus determination module 301 is also used to extract the plaintiff information, the defendant's defense information, and the evidence information from the judgment data; based on the preset dispute focus determination model according to the The plaintiff’s information, the defendant’s defense information, and the evidence information determine a preset number of candidate dispute focus and the output probability value of each candidate dispute focus; the candidate dispute focus with the largest output probability value is taken as the target The focus of the case.

In an embodiment, the calculation module 304 is further configured to obtain first case knowledge information from the first knowledge graph, and obtain respective corresponding second case knowledge information from each of the second knowledge graphs ; According to the first case knowledge information and each of the second case knowledge information, the similarity between the first knowledge graph and each of the second knowledge graphs is calculated.

In an embodiment, the calculation module 304 is further configured to calculate each entity node and each entity in the first knowledge graph according to the first case knowledge information and each of the second case knowledge information. State the target similarity between corresponding entity nodes in the second knowledge graph;

According to the preset coefficient corresponding to each entity node in the first knowledge graph and the similarity of each target, the similarity between the first knowledge graph and each of the second knowledge graphs is calculated.

Please refer to FIG. 5, which is a schematic block diagram of another case recommendation apparatus provided by an embodiment of the application. As shown in FIG. 5, the case recommendation device 400 includes: a dispute focus determination module 401, a graph construction module 402, a first calculation module 403, a candidate case determination module 404, a second calculation module 405, and a case determination module 406.

The dispute focus determination module 401 is configured to obtain the judgment data of the target case, and based on a preset dispute focus determination model, determine the dispute focus of the target case according to the judgment data;

The graph construction module 402 is configured to construct a first knowledge graph of the target case according to the judgment data and the dispute focus;

The first calculation module 403 is used to calculate the similarity between the dispute focus and the dispute focus of each judged case in the pre-stored judged case database;

The candidate case determination module 404 is configured to determine the candidate case set to be recommended from the pre-stored judged case database according to the similarity between the dispute focus and the dispute focus of each judged case;

The second calculation module 405 is further configured to obtain a second knowledge graph corresponding to each candidate case in the candidate case set, and calculate the similarity between the first knowledge graph and each second knowledge graph;

The case determination module 406 is configured to determine the same case and concurrent judgment cases from the set of candidate cases according to the similarity between the first knowledge graph and each of the second knowledge graphs.

In one embodiment, the first calculation module 403 is also used to count the total number of cases that have been judged, and to determine the number of concurrent threads based on the total number of cases; according to the number of concurrent threads, call a preset The corresponding number of idle threads in the thread pool concurrently calculate the similarity between the dispute focus and the dispute focus of each judged case in the pre-stored judged case library.

In an embodiment, the first calculation module 403 is further configured to obtain a mapping relationship table between the number of pre-stored cases and the number of concurrent threads, and query the mapping relationship table to obtain the concurrency corresponding to the total number of cases. Threads.

It should be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the above described device and each module and unit can refer to the corresponding process in the foregoing case recommendation method embodiment. I won't repeat them here.

The apparatus provided in the foregoing embodiment may be implemented in the form of a computer program, and the computer program may run on the computer device as shown in FIG. 6. Please refer to FIG. 6, which is a schematic block diagram of the structure of a computer device provided by an embodiment of the application. The computer device may be a server.

As shown in FIG. 6, the computer device includes a processor, a memory, and a network interface connected through a system bus, where the memory may include a non-volatile storage medium and an internal memory. The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any case recommendation method. The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment. The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can make the processor execute the case recommendation method shown in any of the above embodiments, wherein the Recommended methods for the case include:

According to the judgment data and the dispute focus, construct the first knowledge graph of the target case; according to the dispute focus, determine the candidate case set to be recommended from the pre-stored judged case database;

The network interface is used for network communication, such as sending assigned tasks. Those skilled in the art can understand that the structure shown in FIG. 6 is only a block diagram of part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may Including more or fewer parts than shown in the figure, or combining some parts, or having a different arrangement of parts.

It should be noted that those skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the computer device described above can refer to the corresponding process in the above-mentioned case recommendation method embodiment. Go into details.

The embodiment of the present application also provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, the computer-readable storage medium may be non-volatile or volatile, and the computer The program includes program instructions, and the method implemented when the program instructions are executed can refer to the various embodiments of the case recommendation method of this application. The case recommendation method includes the following steps:

The computer-readable storage medium may be the internal storage unit of the computer device described in the foregoing embodiment, for example, the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk or a smart memory card (Smart Memory Card) equipped on the computer device. Media Card, SMC), Secure Digital (Secure Digital, SD) card, flash memory card (Flash Card) and so on.

Claims

A case recommendation method, which includes:

Obtain the judgment data of the target case, and determine the dispute focus of the target case based on the judgment data based on a preset dispute focus determination model;

Construct a first knowledge graph of the target case according to the judgment data and the dispute focus;

According to the focus of the dispute, determine the candidate case set to be recommended from the pre-stored judged case database;

Acquiring a second knowledge graph corresponding to each candidate case in the candidate case set, and calculating the similarity between the first knowledge graph and each of the second knowledge graphs;

According to the degree of similarity between the first knowledge graph and each of the second knowledge graphs, co-judge cases are determined from the set of candidate cases.
The case recommendation method according to claim 1, wherein the determining the focus of dispute of the target case based on the preset dispute focus determination model and determining the focus of dispute of the target case according to the judgment data comprises:

Extract the plaintiff information, the defendant's defense information and the evidence information from the judgment data;

Based on the preset dispute focus determination model, determine a preset number of candidate dispute focus and the output probability value of each candidate dispute focus according to the plaintiff request information, the defendant's defense information, and the evidence information;

The candidate dispute focus with the largest output probability value is taken as the dispute focus of the target case.
The case recommendation method according to claim 1, wherein the calculating the similarity between the first knowledge graph and each of the second knowledge graphs comprises:

Obtaining first case knowledge information from the first knowledge graph, and obtaining respective corresponding second case knowledge information from each of the second knowledge graphs;

According to the first case knowledge information and each of the second case knowledge information, the similarity between the first knowledge graph and each of the second knowledge graphs is calculated.
The case recommendation method according to claim 3, wherein the first knowledge graph and each second knowledge graph are calculated based on the first case knowledge information and each of the second case knowledge information The similarity between, including:

According to the first case knowledge information and each of the second case knowledge information, the target similarity between each entity node in the first knowledge graph and the corresponding entity node in each second knowledge graph is calculated degree;

According to the preset coefficient corresponding to each entity node in the first knowledge graph and the similarity of each target, the similarity between the first knowledge graph and each of the second knowledge graphs is calculated.
The case recommendation method according to any one of claims 1 to 4, wherein the determining a candidate case set to be recommended from a pre-stored judged case database according to the focus of the dispute comprises:

Calculate the similarity between the dispute focus and the dispute focus of each judged case in the pre-stored judged case database;

According to the similarity between the dispute focus and the dispute focus of each sentenced case, the candidate case set to be recommended is determined from the pre-stored sentenced case database.
The case recommendation method according to claim 5, wherein said calculating the similarity between the focus of dispute and the focus of dispute of each judged case in the pre-stored judged case database comprises:

Count the total number of cases that have been judged, and determine the number of concurrent threads based on the total number of cases;

According to the number of concurrent threads, a corresponding number of idle threads in the preset thread pool are called to concurrently calculate the similarity between the dispute focus and the dispute focus of each judged case in the stored judged case library.
The case recommendation method according to claim 6, wherein the determining the number of concurrent threads according to the total number comprises:

Obtain a mapping relationship table between the number of pre-stored cases and the number of concurrent threads, and query the mapping relationship table to obtain the number of concurrent threads corresponding to the total number of cases.
A case recommendation device, wherein the case recommendation device includes:

The dispute focus determination module is used to obtain the judgment data of the target case, and based on the preset dispute focus determination model, determine the dispute focus of the target case according to the judgment data;

The graph construction module is used to construct the first knowledge graph of the target case according to the judgment data and the dispute focus;

The candidate case determination module is used to determine the candidate case set to be recommended from the pre-stored judged case database according to the focus of the dispute;

A calculation module, configured to obtain a second knowledge graph corresponding to each candidate case in the candidate case set, and calculate the similarity between the first knowledge graph and each of the second knowledge graphs;

The case determination module is configured to determine the same case and concurrent judgment cases from the set of candidate cases according to the similarity between the first knowledge graph and each of the second knowledge graphs.
The case recommendation device according to claim 8, wherein the dispute focus determining module,

It is also used to extract the plaintiff information, the defendant's defense information, and the evidence information from the judgment data;

Based on the preset dispute focus determination model, determine a preset number of candidate dispute focus and the output probability value of each candidate dispute focus according to the plaintiff request information, the defendant's defense information, and the evidence information;

The candidate dispute focus with the largest output probability value is taken as the dispute focus of the target case.
The case recommendation device according to claim 8, wherein the calculation module,

It is also used to obtain first case knowledge information from the first knowledge graph, and obtain respective corresponding second case knowledge information from each of the second knowledge graphs;

According to the first case knowledge information and each of the second case knowledge information, the similarity between the first knowledge graph and each of the second knowledge graphs is calculated.
A computer device, wherein the computer device includes a processor, a memory, and a computer program stored on the memory and executable by the processor, wherein when the computer program is executed by the processor, the A recommended method for a case:

Among them, the recommended method for the case includes:

Obtain the judgment data of the target case, and determine the dispute focus of the target case based on the judgment data based on a preset dispute focus determination model;

Construct a first knowledge graph of the target case according to the judgment data and the dispute focus;

According to the focus of the dispute, determine the candidate case set to be recommended from the pre-stored judged case database;

Acquiring a second knowledge graph corresponding to each candidate case in the candidate case set, and calculating the similarity between the first knowledge graph and each of the second knowledge graphs;

According to the degree of similarity between the first knowledge graph and each of the second knowledge graphs, co-judge cases are determined from the set of candidate cases.
11. The computer device according to claim 11, wherein the determining the focus of dispute of the target case based on the preset dispute focus determination model according to the judgment data comprises:

Extract the plaintiff information, the defendant's defense information and the evidence information from the judgment data;

Based on the preset dispute focus determination model, determine a preset number of candidate dispute focuses and the output probability value of each candidate dispute focus according to the plaintiff information, the defendant's defense information, and the evidence information;

The candidate dispute focus with the largest output probability value is taken as the dispute focus of the target case.
The computer device according to claim 11, wherein the calculating the similarity between the first knowledge graph and each of the second knowledge graphs comprises:

Obtaining first case knowledge information from the first knowledge graph, and obtaining respective corresponding second case knowledge information from each of the second knowledge graphs;

According to the first case knowledge information and each of the second case knowledge information, the similarity between the first knowledge graph and each of the second knowledge graphs is calculated.
The computer device according to claim 13, wherein the calculation of the first knowledge graph and each of the second knowledge graphs is based on the first case knowledge information and each of the second case knowledge information The similarity between:

According to the first case knowledge information and each of the second case knowledge information, the target similarity between each entity node in the first knowledge graph and the corresponding entity node in each second knowledge graph is calculated degree;

According to the preset coefficient corresponding to each entity node in the first knowledge graph and the similarity of each target, the similarity between the first knowledge graph and each of the second knowledge graphs is calculated.
The computer device according to claims 11 to 14, wherein the determining the candidate case set to be recommended from the pre-stored judged case database according to the dispute focus includes:

Calculate the similarity between the dispute focus and the dispute focus of each judged case in the pre-stored judged case database;

According to the similarity between the dispute focus and the dispute focus of each sentenced case, the candidate case set to be recommended is determined from the pre-stored sentenced case database.
A computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, a case recommendation method is implemented, wherein the case recommendation method includes the following step:

Obtain the judgment data of the target case, and determine the dispute focus of the target case based on the judgment data based on a preset dispute focus determination model;

Construct a first knowledge graph of the target case according to the judgment data and the dispute focus;

According to the focus of the dispute, determine the candidate case set to be recommended from the pre-stored judged case database;

Acquiring a second knowledge graph corresponding to each candidate case in the candidate case set, and calculating the similarity between the first knowledge graph and each of the second knowledge graphs;

According to the degree of similarity between the first knowledge graph and each of the second knowledge graphs, co-judge cases are determined from the set of candidate cases.
The computer-readable storage medium according to claim 16, wherein the determining the focus of dispute of the target case based on the preset dispute focus determination model according to the judgment data comprises:

Extract the plaintiff information, the defendant's defense information and the evidence information from the judgment data;

Based on the preset dispute focus determination model, determine a preset number of candidate dispute focus and the output probability value of each candidate dispute focus according to the plaintiff request information, the defendant's defense information, and the evidence information;

The candidate dispute focus with the largest output probability value is taken as the dispute focus of the target case.
The computer-readable storage medium according to claim 16, wherein the calculating the similarity between the first knowledge graph and each of the second knowledge graphs comprises:

Obtaining first case knowledge information from the first knowledge graph, and obtaining respective corresponding second case knowledge information from each of the second knowledge graphs;

According to the first case knowledge information and each of the second case knowledge information, the similarity between the first knowledge graph and each of the second knowledge graphs is calculated.
18. The computer-readable storage medium of claim 18, wherein the first knowledge graph and each second case knowledge information are calculated based on the first case knowledge information and each of the second case knowledge information. The similarity between knowledge graphs, including:

According to the first case knowledge information and each of the second case knowledge information, the target similarity between each entity node in the first knowledge graph and the corresponding entity node in each second knowledge graph is calculated degree;

According to the preset coefficient corresponding to each entity node in the first knowledge graph and the similarity of each target, the similarity between the first knowledge graph and each of the second knowledge graphs is calculated.
The computer-readable storage medium according to claims 16 to 19, wherein the determining the candidate case set to be recommended from the pre-stored judged case database according to the dispute focus includes:

Calculate the similarity between the dispute focus and the dispute focus of each judged case in the pre-stored judged case database;

According to the similarity between the dispute focus and the dispute focus of each sentenced case, the candidate case set to be recommended is determined from the pre-stored sentenced case database.