Case relationship network construction method, model training method and case processing method

Abstract

The application provides a case relationship network construction method, a model training method and a case processing method. The case relationship network construction method comprises the following steps: for a case in a case library, extracting a target text in the case, the target text being used to describe the basic situation of the case; performing classification processing on the target text to obtain an event contained in the case; constructing a case line corresponding to the case in the case relationship network according to the event, the case line comprising an identification point of the corresponding event; and constructing a case relationship network corresponding to the case library according to the case line corresponding to the case in the case library, wherein the case lines with the same event in the case relationship network intersect at the identification point corresponding to the same event, and the case relationship network is used for the determination of similar cases. The application can realize the accurate determination of similar cases.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to methods for constructing case relationship networks, model training methods, and case processing methods. Background Technology

[0002] A case comprises one or more events. When handling a case, users refer to the handling process of similar cases. However, when there are many cases in the case database, it takes a lot of time for users to accurately locate similar cases, resulting in low efficiency in identifying similar cases.

[0003] Currently, methods for identifying similar cases, such as TF-IDF (term frequency–inverse document frequency), are based on the word frequencies in the entire case and represent the case with discrete vectors. However, the text length of cases is usually very large, and discrete vectors cannot accurately represent such long texts, thus making it impossible to accurately identify similar cases. Summary of the Invention

[0004] This application provides methods for constructing case relationship networks, model training, and case processing to accurately identify similar cases.

[0005] The first aspect of this application provides a method for constructing a case relationship network, comprising: extracting target text from cases in a case database, the target text describing the basic situation of the case; classifying the target text to obtain the events contained in the case; constructing case lines corresponding to the cases in the case relationship network based on the events, the case lines including the marker points of the corresponding events; and constructing a case relationship network corresponding to the case database based on the case lines corresponding to the cases in the case database, wherein case lines with the same events in the case relationship network intersect at the marker points corresponding to the same events, and the case relationship network is used to determine similar cases.

[0006] The second aspect of this application provides a model training method, including: obtaining a case relationship network corresponding to a case database, wherein the case relationship network is determined according to the case relationship network construction method of the first aspect; training a case processing model using the case relationship network; and obtaining a trained case processing model when the case processing model meets preset training requirements, wherein the case processing model is used for retrieval of similar cases or clustering of similar cases.

[0007] The third aspect of this application provides a case processing method, comprising: extracting target text of a case to be processed, the target text being used to describe the basic situation of the case to be processed; classifying the target text to obtain the events contained in the case to be processed; inputting the events contained in the case to be processed into a case processing model for processing to obtain similar cases of the case to be processed, wherein the case processing model is trained according to the model training method of the second aspect.

[0008] The fourth aspect of this application provides a case processing method, including: acquiring a case to be processed; sending the case to be processed to a server; and receiving similar cases of the case to be processed sent by the server, wherein the similar cases are obtained according to the case processing method of the third aspect.

[0009] A fifth aspect of this application provides a case relationship network construction apparatus, comprising:

[0010] The extraction module is used to extract target text from cases in the case database. The target text is used to describe the basic information of the case.

[0011] The classification module is used to classify the target text to obtain the events contained in the case;

[0012] The first construction module is used to construct the case line corresponding to the case in the case relationship network based on the event. The case line includes the marker point of the corresponding event.

[0013] The second construction module is used to construct a case relationship network corresponding to the case database based on the case lines corresponding to the cases in the case database. Case lines with the same event in the case relationship network intersect at the markers corresponding to the same event. The case relationship network is used to determine similar cases.

[0014] A sixth aspect of this application provides a case processing system, including: a cloud server and a terminal device, wherein a case processing model is deployed on the cloud server;

[0015] Terminal devices are used to acquire pending cases and send them to the cloud server.

[0016] The cloud server is used to extract the target text of the pending case, which describes the basic situation of the pending case; classify the target text to obtain at least one event of the pending case; input the at least one event into the case processing model for processing to obtain similar cases of the pending case, the case processing model is trained according to the model training method of the second aspect; and send the similar cases to the terminal device.

[0017] A seventh aspect of this application provides an electronic device, including: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a case relationship network construction method as described in the first aspect, and / or a model training method as described in the second aspect, and / or a case processing method as described in the third or fourth aspect.

[0018] The eighth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the model pre-training method of the first aspect, and / or the model training method of the second aspect, and / or the case processing method of the third aspect.

[0019] This application embodiment is applied to the scenario of determining similar cases. It extracts target text from cases in a case database, the target text describing the basic situation of the case; categorizes the target text to obtain the events contained in the case; based on the events, constructs case lines corresponding to the cases in a case relationship network, the case lines including the marker points of the corresponding events; based on the case lines corresponding to cases in the case database, constructs a case relationship network corresponding to the case database, wherein case lines with the same events in the case relationship network intersect at the marker points corresponding to the same events. The case relationship network is used to determine similar cases, thereby achieving accurate determination of similar cases. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0021] Figure 1 An application scenario diagram provided for an exemplary embodiment of this application;

[0022] Figure 2 A flowchart illustrating the steps of a case relationship network construction method provided as an exemplary embodiment of this application;

[0023] Figure 3 A schematic diagram of case lines in a case relationship network provided for an exemplary embodiment of this application;

[0024] Figure 4 A schematic diagram of a case relationship network provided for an exemplary embodiment of this application;

[0025] Figure 5 A flowchart illustrating the steps of a model training method provided in an exemplary embodiment of this application;

[0026] Figure 6 A flowchart illustrating the steps of a case processing method provided as an exemplary embodiment of this application;

[0027] Figure 7 A flowchart illustrating the steps of another case processing method provided as an exemplary embodiment of this application;

[0028] Figure 8 A structural block diagram of a case relationship network construction device provided for an exemplary embodiment of this application;

[0029] Figure 9 This is a schematic diagram of the structure of an electronic device provided as an exemplary embodiment of this application. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0031] In related technologies, knowledge graphs are used to identify relationships between objects. A knowledge hypergraph is composed of a structure of (first node, edges, and last node), where edges connect the first and last nodes, representing the relationship between them. However, for case databases containing a large number of cases, the relationships between cases are often unclear, and knowledge graphs using these technologies cannot represent these relationships. Therefore, a method to represent the relationships between cases is urgently needed.

[0032] To address the aforementioned issues, this application provides a method for constructing a case relationship network, applied to the scenario of identifying similar cases. The method involves extracting target text from cases in a case database, where the target text describes the basic situation of the case; classifying the target text to obtain the events contained in the case; constructing case lines corresponding to the cases in the case relationship network based on the events, where each case line includes markers for the corresponding events; and constructing a case relationship network corresponding to the case database based on the case lines corresponding to the cases in the case database, where case lines with the same events intersect at markers corresponding to the same events. This case relationship network is used to identify similar cases. Based on at least one event contained in a case, a case can be represented as a case line. The interactions between case lines can further model the interaction relationships between cases, thereby helping to better understand cases and identify similar cases.

[0033] In this embodiment, the execution device of the case relationship network construction method is not limited. Optionally, the case relationship network construction method can be implemented using a cloud computing system. For example, the case relationship network construction method can be applied to a cloud server to leverage the advantages of cloud resources to run various neural network models; rather than being applied to the cloud, the case relationship network construction method can also be applied to server-side devices such as conventional servers, cloud servers, or server arrays.

[0034] In addition, refer to Figure 1 This diagram illustrates one application scenario of this application. It includes a case processing system comprising a server and terminal devices, the server being a cloud server. Within the server, a case relationship network of a case database can be constructed, and a case processing model can be trained. The trained case processing model is then uploaded to the server. When the server receives a case to be processed from a terminal device, it can use the case processing model to determine similar cases to that case.

[0035] Specifically, a case processing system is provided, comprising: a cloud server and terminal devices; a case processing model deployed on the cloud server; terminal devices for acquiring cases to be processed and sending them to the cloud server; the cloud server for extracting target text from the cases to be processed, the target text describing the basic information of the cases to be processed; classifying the target text to obtain at least one event of the cases to be processed; inputting the at least one event into the case processing model for processing to obtain similar cases of the cases to be processed, the case processing model being trained according to a model training method; and sending similar cases to the terminal devices.

[0036] Figure 1 This is merely one example of an application scenario for this application. This application can also be applied to other case handling scenarios, which are not limited here.

[0037] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0038] Figure 2 A flowchart illustrating the steps of a case relationship network construction method provided for an exemplary embodiment of this application. Figure 2 The method for constructing the relationship network in this case, as shown, specifically includes the following steps:

[0039] S201, Extract the target text from the cases in the case database.

[0040] The target text is used to describe the basic facts of the case.

[0041] In this application, a case is a description of at least one event that occurs under specific conditions, such as a legal case or a news case. The text of a case is generally quite long, and this application can first extract the "basic facts" of the case as the target text.

[0042] S202, classify the target text to obtain the events contained in the case.

[0043] The process of classifying the target text to obtain the events contained in the case includes: inputting the target text into a pre-trained event classification model for classification processing to obtain the events contained in the case.

[0044] In this application, the event classification model can be a neural network model employing BERT and CRF (Conditional Random Field). BERT is a large-scale, trained language model. CRF is used for information extraction. Specifically, by inputting the target text into the event classification model for event classification, it can be determined that a case contains at least one event. Events are a crucial dimension for characterizing a case; using events to represent a case is more reasonable and accurate than using statistical information such as TF-IDF, word frequency, or document length.

[0045] Furthermore, the effectiveness of event extraction affects the construction of the case relationship network. The event classification model uses BERT and CRF as its most basic models. The RoBERTa-wwm-ext pre-trained model can also be used to improve the F1 score of event extraction. A simple hypergraph cannot represent event types, and a simple knowledge graph cannot model multiple events simultaneously within a single case line. The case line in the case relationship network provides a new solution. Knowledge hyper-case lines can be expressed in various ways, and the (event:event type) pair format adopted in this application is more in line with legal scenarios. Specifically, RoBERTa-wwm-ext includes the RoBERTa (A Robustly Optimized BERT) model and wwm (Whole Word Mask).

[0046] Furthermore, for legal cases, given their highly specialized nature, pre-defining legal events is crucial for characterizing case features. The LEVEN dataset (a large-scale legal event detection dataset) defines 108 fine-grained events, including criminal, civil, general conduct, and prohibited conduct, along with 18 corresponding event types. This step primarily involves identifying the events contained in the case based on the target text, and then determining the event type based on the correspondence between events and event types.

[0047] For example, for an event type of "property infringement", the corresponding events include robbery, property damage, property loss, and theft.

[0048] In this embodiment of the application, events are obtained by classifying the target text, thereby compressing the target text and saving a large amount of computing space required for processing long texts.

[0049] S203, Based on the event, construct the case line corresponding to the case in the case relationship network.

[0050] The case line includes markers for corresponding events. There is a one-to-one correspondence between events and markers, with one marker representing one event. Therefore, the number of events contained in a case is equal to the number of markers.

[0051] Furthermore, based on the events, a case line corresponding to the case in the case relationship network is constructed, including: determining the event type of the event; and constructing the case line of the case in the case relationship network based on the events and event types, wherein the marker points also reflect the event type of the event. In this application, a correspondence between events and event types is pre-set, with one event having one event type and one event type corresponding to multiple events.

[0052] In this embodiment, events or (events, event types) can be organized and associated to form case lines in a case relationship network. It is evident that each case line in the case relationship network corresponds one-to-one with a case. Specifically, the case relationship network is a combination of a hypergraph and a knowledge graph. It inherits the characteristic of a hypergraph where a case line contains multiple identifiers; in a hypergraph, a case line is represented as (node ​​1, node 2, ..., node n). It also inherits the multi-relationship characteristic of a knowledge graph; in a knowledge graph, a case line is represented as (head entity, relationship, tail entity). Entities are connected by relationships, which enrich the information of the case lines in the graph structure. In the case relationship network structure, a case line is represented as: Case: (Event, Event Type 1: Event 2, Event Type 2:, ..., Event n, Event Type n:)

[0053] Reference Figure 3 A case line represents a single case, containing markers 1 through 5. Each marker represents a corresponding event and / or (event and event type). For example, marker 1 represents event 1 and event type A, marker 2 represents event 2 and event type B, marker 3 represents event 3 and event type A, marker 4 represents event 4 and event type C, and marker 5 represents event 5 and event type C.

[0054] S204. Based on the case lines corresponding to the cases in the case database, construct the case relationship network corresponding to the case database.

[0055] In this embodiment of the application, a case relationship network can be completed by modeling the case line for each case in the case database. Specifically, each case is constructed with knowledge super-case lines, and all knowledge super-case lines will form a larger case relationship network. More in-depth mining can be carried out on the case relationship network, such as the retrieval of similar cases and / or the clustering of similar cases.

[0056] In the case relationship network, case lines with the same event intersect at the markers corresponding to the same event. The case relationship network is used to identify similar cases.

[0057] For example, refer to Figure 4 Case line (marker 1, marker 2, marker 3, marker 4, marker 5) represents case A. Case line (marker 9, marker 7, marker 10) represents case B. Case line (marker 6, marker 2, marker 7, marker 8) represents case C. Case line (marker 15, marker 14, marker 13, marker 2) represents case D. Case line (marker 12, marker 4, marker 8, marker 11) represents case E.

[0058] In the embodiments of this application, the case line can be a straight line, a curve, or a broken line, and is not limited thereto.

[0059] Furthermore, the case relationship network in this application is an extension of the knowledge graph. In a knowledge graph, a single piece of data is usually represented in the form of a head entity, a relation, and a tail entity. That is, a piece of data contains two identifiers and one relation. However, a piece of data in the case relationship network can contain multiple identifiers (more than two identifiers) and multiple relations (more than one relation point).

[0060] This application provides a method for constructing a case relationship network, in which events are a crucial dimension for characterizing a case. Using events to represent a case offers greater interpretability than using statistical information such as TF-IDF, word frequency, or document length. This application uses events within a case to model it. Furthermore, this application models cases as knowledge case lines and the entire case database as a case relationship network, representing a practical application of case relationship networks in real-world scenarios. Using this case relationship network construction method, users (such as judges or lawyers and other legal professionals) can more easily find the connections between cases based on events, leading to a more accurate and profound understanding of the cases to be handled.

[0061] Furthermore, the case relationship network construction method provided in this application can solve the problem of how to represent legal cases, providing a new solution for legal case retrieval and data support for legal case association mining. This application first extracts events from legal cases, then determines the event type corresponding to each event, and then constructs a knowledge case line based on the event and event type. Connecting all the knowledge case lines corresponding to cases in the case database further forms a larger-scale case relationship network. Further case retrieval, case clustering, and other subsequent tasks can be performed on this case-aware case relationship network. First, the key elements of events and event types are extracted through an event classification model. Events can represent fine-grained information about a case, while event types can represent coarse-grained information. Compared to other case representation methods, this method can capture case information and is event-aware.

[0062] This application represents a case as a hyper-case line, with each hyper-case line containing identifiers for multiple cases. Based on this, the entire case database can be represented as a case relationship network. Various data mining operations can be performed on this case relationship network, offering greater scalability compared to traditional frequency-based statistical methods. With this method for constructing case relationship networks in legal cases, judges, lawyers, and other legal professionals can more easily find connections between cases based on events, leading to a more accurate and profound understanding of the cases.

[0063] Reference Figure 5 The above is a flowchart illustrating the steps of a model training method provided as an exemplary embodiment of this application. Figure 5 The model training method shown includes the following steps:

[0064] S501, obtain the case relationship network corresponding to the case database.

[0065] The case relationship network is determined based on the aforementioned case relationship network construction method, and will not be elaborated further here.

[0066] S502 uses a case relationship network to train a case processing model.

[0067] Specifically, similar case retrieval involves inputting a case into the case processing model, which then outputs similar cases. Similar case clustering involves inputting multiple cases into the case processing model, which categorizes these cases, classifying cases into the same category as similar cases. Alternatively, inputting a case into the model outputs its category, allowing users to search for similar cases belonging to the same category.

[0068] The process of training a case processing model using a case relationship network includes: inputting case lines from the case relationship network into the encoder of the case processing model for encoding to obtain the encoding vectors corresponding to the case lines; determining the similarity between the encoding vectors of two case lines in the normalization module of the case processing model; determining the loss value between the similarity and the ground truth value of the similarity between the two case lines; and adjusting the encoder based on the loss value to obtain the trained case processing model.

[0069] This application trains a case processing model, enabling it to learn about the case relationship network and accurately process cases. Specifically, inputting the case line into the encoder of the case processing model for encoding means inputting the identifier points of the case line into the encoder for encoding. For example, inputting the case line (event, event type 1: event 2, event type 2:, ..., event n, event type n) into the encoder for encoding results in the encoded vector of the case line, which is essentially the encoded vector of (event, event type 1: event 2, event type 2:, ..., event n, event type n), and this encoded vector is a continuous vector.

[0070] Furthermore, the similarity between two case lines can be pre-determined as the ground truth value. If the cases represented by the two case lines are similar, the ground truth value is 1. If the cases represented by the two case lines are dissimilar, the ground truth value is 0. A normalization module can then be used to determine the similarity between the encoded vectors corresponding to the two case lines using a softmax normalization method. The loss value between this similarity and the ground truth value is then calculated. If the loss value is greater than a threshold, it can be used to adjust the encoder's encoding parameters.

[0071] S503, if the case handling model meets the preset training requirements, the trained case handling model is obtained.

[0072] Among them, the case processing model is used for retrieving similar cases or clustering similar cases.

[0073] In this embodiment of the application, when the loss value is less than the loss value threshold, it can be determined that the case processing model meets the preset training requirements.

[0074] Furthermore, in this embodiment of the application, the cases in the case database can be cases that have already been processed. When a user receives a new case to be processed, the user can use the case processing model to retrieve similar cases from the case database for reference in processing the case to be processed.

[0075] In this embodiment of the application, by training the case processing model using a case relationship network, the case processing model can learn the relationship between various cases, more accurately represent the cases, and retrieve similar cases.

[0076] Figure 6 A flowchart illustrating the steps of a case processing method provided for an exemplary embodiment of this application. Figure 6 The case handling method shown includes the following steps:

[0077] S601, Extract the target text of the case to be processed.

[0078] The target text is used to describe the basic information of the case to be processed.

[0079] S602, classify the target text to obtain the events contained in the case to be processed.

[0080] S603: Input the events contained in the case to be processed into the case processing model for processing to obtain similar cases to the case to be processed.

[0081] The case processing model was trained using the aforementioned model training method.

[0082] Furthermore, the events contained in the case to be processed are input into the case processing model for processing to obtain similar cases of the case to be processed, including: determining the event type corresponding to the event; inputting the event and the event type corresponding to the event into the case processing model for processing to obtain similar cases of the case to be processed.

[0083] Specific details regarding the handling of this case are described above and will not be repeated here.

[0084] This application can be applied in the legal industry. When legal professionals take on a case, they often refer to similar past cases. The judgments of past cases are of significant reference value for the judgment of new cases. In the past, judges or lawyers usually had to manually search for similar cases in a large number of case files or simply search by keywords, which was inefficient. This application can improve the effectiveness of the search, provide legal professionals with more accurate search results, and improve judicial efficiency.

[0085] Figure 7 A flowchart illustrating the steps of a case processing method provided for an exemplary embodiment of this application. Figure 7 The case processing method shown, applied to terminal devices, specifically includes the following steps:

[0086] S701, retrieve pending cases.

[0087] The pending case can be a legal case, or an unresolved case.

[0088] S702, send pending cases to the server.

[0089] The server can use the above-mentioned case processing method to extract the target text of the case to be processed; classify the target text to obtain the events contained in the case to be processed; input the events contained in the case to be processed into the case processing model for processing to obtain similar cases of the case to be processed.

[0090] S703, Receive similar cases to pending cases sent by the server.

[0091] The similar cases were obtained based on the aforementioned case handling methods.

[0092] For details regarding the specific methods used to handle this case, please refer to the description provided by Sinopec; these details will not be repeated here.

[0093] This application can be applied in the legal industry. When legal professionals take on a case, they often refer to similar past cases. The judgments of past cases are of significant reference value for the judgment of new cases. In the past, judges or lawyers usually had to manually search for similar cases in a large number of case files or simply search by keywords, which was inefficient. This application can improve the effectiveness of the search, provide legal professionals with more accurate search results, and improve judicial efficiency.

[0094] In this application embodiment, in addition to providing a method for constructing a case relationship network, a device for constructing a case relationship network is also provided, such as... Figure 8 As shown, the case relationship network construction device 80 includes:

[0095] Extraction module 81 is used to extract target text from cases in the case database. The target text is used to describe the basic information of the case.

[0096] Classification module 82 is used to classify the target text to obtain the events contained in the case;

[0097] The first construction module 83 is used to construct the case line corresponding to the case in the case relationship network based on the event. The case line includes the marker point of the corresponding event.

[0098] The second construction module 84 is used to construct a case relationship network corresponding to the case database based on the case lines corresponding to the cases in the case database. Case lines with the same event in the case relationship network intersect at the marker points corresponding to the same event. The case relationship network is used to determine similar cases.

[0099] In an optional embodiment, the first construction module 83 is specifically used to determine the event type of the event; and to construct a case line in the case relationship network based on the event and the event type, wherein the marker point also reflects the event type of the event.

[0100] In an optional embodiment, the classification module 82 is specifically used to classify the target text input into a pre-trained event classification model to obtain the events contained in the case.

[0101] The case relationship network construction apparatus provided in this application embodiment can improve the accuracy of identifying similar cases. The specific implementation process is the same as described in the above method embodiment, and will not be repeated here.

[0102] In addition, this application also provides a model training apparatus (not shown), comprising:

[0103] The acquisition module is used to acquire the case relationship network corresponding to the case database. The case relationship network is determined by any of the case relationship network construction methods mentioned above.

[0104] The training module is used to train the case processing model using a case relationship network;

[0105] The module determines the case processing model if it meets the preset training requirements, and then obtains the trained case processing model. The case processing model is used for the retrieval of similar cases or the clustering of similar cases.

[0106] In an optional embodiment, the training module is specifically used for:

[0107] The case lines in the case relationship network are input into the encoder in the case processing model for encoding processing to obtain the encoding vectors corresponding to the case lines.

[0108] In the normalization module of the case processing model, the similarity of the encoding vectors of two case lines is determined;

[0109] Determine the loss value for similarity and the true value of similarity between two case lines;

[0110] The encoder is adjusted based on the loss value to obtain the trained case processing model.

[0111] In addition, this application also provides a case processing device (not shown), comprising:

[0112] The extraction module is used to extract the target text of the case to be processed. The target text describes the basic information of the case to be processed.

[0113] The processing module is used to classify the target text to obtain the events contained in the case to be processed; the events contained in the case to be processed are input into the case processing model for processing to obtain similar cases of the case to be processed. The case processing model is trained according to the model training method described above.

[0114] In one optional embodiment, the processing module is specifically used to: determine the event type corresponding to the event; process the event and the event type corresponding to the event into the processing model to obtain similar cases of the case to be processed.

[0115] In addition, this application also provides another case processing device (not shown), applied to a terminal device, including:

[0116] The acquisition module is used to acquire cases to be processed.

[0117] The sending module is used by the server to send pending cases;

[0118] The receiving module is used to receive similar cases to the pending cases sent by the server. The similar cases are obtained according to the case processing method described above.

[0119] Furthermore, in some of the processes described in the above embodiments and accompanying drawings, multiple operations appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The sequence numbers are merely used to distinguish different operations, and the sequence numbers themselves do not represent any execution order. Additionally, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0120] Figure 9 This is a schematic diagram of an electronic device provided as an exemplary embodiment of this application. The electronic device can be a cloud device. This device is used to run the aforementioned model training method and model pre-training method. Figure 9 As shown, the electronic device includes a memory 94 and a processor 95.

[0121] Memory 94 is used to store computer programs and can be configured to store various other data to support operation on electronic devices. This memory 94 may be object storage (OSS).

[0122] The memory 94 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0123] Processor 95, coupled to memory 94, is used to execute computer programs in memory 94 for: extracting target text from cases in a case database, the target text describing the basic facts of the case; classifying the target text to obtain the events contained in the case; constructing case lines corresponding to the cases in a case relationship network based on the events, the case lines including the marker points of the corresponding events; and constructing a case relationship network corresponding to the case database based on the case lines corresponding to the cases in the case database, wherein case lines with the same events in the case relationship network intersect at the marker points corresponding to the same events, and the case relationship network is used to determine similar cases.

[0124] Further optionally, when the processor 95 constructs the case line corresponding to the case in the case relationship network based on the event, it is specifically used to: determine the event type of the event; construct the case line of the case in the case relationship network based on the event and the event type, wherein the marker point also reflects the event type of the event.

[0125] Further optionally, when the processor 95 performs classification processing on the target text to obtain the events contained in the case, it is specifically used to: input the target text into a pre-trained event classification model for classification processing to obtain the events contained in the case.

[0126] In one optional embodiment, the processor 95, coupled to the memory 94, is used to execute a computer program in the memory 94, and is further used to: obtain a case relationship network corresponding to the case database, the case relationship network being determined according to the case relationship network construction method of any of the above; train a case processing model using the case relationship network, and, if the case processing model meets the preset training requirements, obtain a trained case processing model, the case processing model being used for the retrieval of similar cases or the clustering of similar cases.

[0127] Further optionally, when the processor 95 trains the case processing model using the case relationship network, it specifically performs the following: inputting the case lines in the case relationship network into the encoder of the case processing model for encoding processing to obtain the encoding vectors corresponding to the case lines; determining the similarity between the encoding vectors of two case lines in the normalization module of the case processing model; determining the loss value between the similarity and the true value of the similarity between the two case lines; and adjusting the encoder according to the loss value to obtain the trained case processing model.

[0128] In one optional embodiment, the processor 95, coupled to the memory 94, is used to execute a computer program in the memory 94, and is further used to: extract target text of a case to be processed, the target text being used to describe the basic situation of the case to be processed; classify the target text to obtain the events contained in the case to be processed; input the events contained in the case to be processed into a case processing model for processing to obtain similar cases of the case to be processed, the case processing model being trained according to the above-described model training method.

[0129] Further optionally, when the processor 95 inputs the events contained in the case to be processed into the case processing model for processing to obtain similar cases of the case to be processed, it is specifically used to: determine the event type corresponding to the event; input the event and the event type corresponding to the event into the case processing model for processing to obtain similar cases of the case to be processed.

[0130] Furthermore, such as Figure 9 As shown, the electronic device also includes other components such as a firewall 91, a load balancer 92, a communication component 96, and a power supply component 98. Figure 9 The diagram only shows some components and does not mean that the electronic device includes only these components. Figure 9 The components shown.

[0131] Accordingly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to implement the steps in the method described above.

[0132] Accordingly, embodiments of this application also provide a computer program product, including a computer program / instructions, which, when executed by a processor, cause the processor to implement the steps in the method described above.

[0133] The above Figure 9 The communication component is configured to facilitate wired or wireless communication between the device containing the communication component and other devices. The device containing the communication component can access wireless networks based on communication standards, such as WiFi, 2G, 3G, 4G / LTE, 5G, or combinations thereof. In one exemplary embodiment, the communication component receives broadcast signals or broadcast-related text from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the communication component also includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID), Infrared Data Association (IrDA) technology, Ultra-Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0134] The above Figure 9 The power supply component provides power to the various components of the device in which it resides. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device in which it resides.

[0135] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0136] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0137] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0138] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0139] In a typical configuration, a computing device includes one or more processors (CPU and / or GPU), input / output interfaces, network interfaces, and memory.

[0140] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0141] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can be used to store text by any method or technology. Text can be computer-readable instructions, data structures, modules of programs, 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 technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store text accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0142] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0143] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A model training method, characterized in that, include: Obtain the case relationship network corresponding to the case database; In the case relationship network, case lines with the same event intersect at the corresponding marker point of the same event. The case relationship network is used to determine similar cases. The event is obtained based on the target text of the case in the case database. The target text is used to describe the case. The case processing model is trained using the aforementioned case relationship network; If the case processing model meets the preset training requirements, a trained case processing model is obtained. The case processing model is used for the retrieval of similar cases or the clustering of similar cases. The step of training the case processing model using the case relationship network includes: The case lines in the case relationship network are input into the encoder in the case processing model for encoding processing to obtain the encoding vectors corresponding to the case lines. In the normalization module of the case processing model, the similarity of the encoding vectors of two case lines is determined; Determine the loss value of the similarity and the true value of the similarity between the two case lines; The encoder is adjusted based on the loss value.

2. The model training method according to claim 1, characterized in that, The case relationship network was constructed using the following method: Extract the target text from the cases in the case database; The target text is categorized to obtain the events contained in the case; Based on the event, construct the case line corresponding to the case in the case relationship network, and the case line includes the marker point corresponding to the event; Based on the case lines corresponding to the cases in the case database, construct the case relationship network corresponding to the case database.

3. The model training method according to claim 2, characterized in that, The step of constructing the case line corresponding to the case in the case relationship network based on the event includes: Determine the event type of the event; Based on the event and the event type, a case line is constructed in the case relationship network for the case, wherein the marker point also reflects the event type of the event.

4. The model training method according to claim 2 or 3, characterized in that, The process of classifying the target text to obtain the events contained in the case includes: The target text is input into a pre-trained event classification model for classification processing to obtain the events contained in the case.

5. A case handling method, characterized in that, include: Extract the target text of the case to be processed, the target text being used to describe the situation of the case to be processed; The target text is categorized to obtain the events contained in the case to be processed; The events contained in the case to be processed are input into the case processing model for processing to obtain similar cases of the case to be processed. The case processing model is trained by the model training method according to claim 1.

6. The case handling method according to claim 5, characterized in that, The process of inputting the events contained in the case to be processed into the case processing model for processing, and obtaining similar cases of the case to be processed, includes: Determine the event type corresponding to the event; The event and its corresponding event type input model are processed to obtain similar cases to the case to be processed.

7. A case handling method, characterized in that, Applied to terminal devices, including: Obtain pending cases; Send the pending case to the server; The server receives similar cases to the pending case, wherein the similar cases are obtained by the case processing method according to claim 5 or 6.

8. A case processing system, characterized in that, include: Cloud server and terminal equipment, wherein a case processing model is deployed on the cloud server; The terminal device is used to acquire cases to be processed and send the cases to be processed to the cloud server; The cloud server is used to extract the target text of the pending case, and the target text is used to describe the situation of the pending case. The target text is categorized to obtain at least one event of the case to be processed; The at least one event is input into the case processing model for processing to obtain similar cases of the case to be processed, wherein the case processing model is trained according to the model training method according to claim 1; and the similar cases are sent to the terminal device.

9. The case processing system according to claim 8, characterized in that, The case processing model is trained using a case relationship network, which is obtained through a case relationship network construction device, which includes: The extraction module is used to extract target text from cases in the case database, wherein the target text is used to describe the circumstances of the case; The classification module is used to classify the target text to obtain the events contained in the case; The first construction module is used to construct a case line corresponding to the case in the case relationship network based on the event, wherein the case line includes an identifier point corresponding to the event; The second construction module is used to construct a case relationship network corresponding to the case database based on the case lines corresponding to the cases in the case database. In the case relationship network, case lines with the same event intersect at the marker points corresponding to the same event. The case relationship network is used to determine similar cases.

10. An electronic device, characterized in that, include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the model training method as described in any one of claims 1 to 4, and / or the case processing method as described in any one of claims 5 to 7.

Images