Case Processing Method, Device, and Storage Medium Based on Knowledge Graph

Abstract

The present disclosure relates to the field of computer technology, and relates to a case processing method and apparatus, a storage medium, and an electronic device based on a knowledge graph. The method includes: extracting target keywords from current case data, matching the target keywords with a pre-constructed knowledge graph to obtain a first vector representation of the current case; inputting the target keywords into a pre-constructed knowledge graph case index, searching to obtain candidate matching cases of the current case and corresponding second vector representations, wherein vector representations corresponding to the cases in the case index have been pre-stored; based on a preset similarity model, obtaining the similarity between the first vector representation and each of the second vector representations, and determining a set of matching cases from the candidate matching cases according to the similarity, so as to process the current case based on the set of matching cases. The present disclosure can improve the accuracy of case retrieval and ensure the efficient execution of subsequent business processes.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and more particularly, to a case processing method based on a knowledge graph, a case processing apparatus based on a knowledge graph, a computer storage medium, and an electronic device. Background Art

[0002] With the development of the field of computer technology, many industries have gradually replaced manual work with computers to achieve efficient automated business processing processes. In many scenarios, computers can also assist humans to complete work, such as assisting users in indexing and searching for relevant case information so that users can efficiently complete tasks based on the search content.

[0003] In the related art, the retrieval of case information is mainly through keyword matching. The retrieval request is segmented and retrieved in the established index, and the retrieval results are scored, sorted, and fed back to the user. However, this method lacks semantic-level matching of cases, resulting in low retrieval accuracy and affecting business processing efficiency.

[0004] It should be noted that the information disclosed in the above background art is only used to enhance the understanding of the background of the present disclosure, and thus may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] An object of the present disclosure is to provide a case processing method and apparatus, a computer storage medium, and an electronic device based on a knowledge graph, thereby at least to a certain extent avoiding problems such as affecting business processing efficiency due to low case retrieval accuracy.

[0006] Other features and advantages of the present disclosure will become apparent through the following detailed description, or be learned in part through the practice of the present disclosure.

[0007] According to an aspect of the present disclosure, there is provided a case processing method based on a knowledge graph, including: extracting target keywords from current case data, matching the target keywords with a pre-constructed knowledge graph to obtain a first vector representation of the current case; inputting the target keywords into a pre-constructed case index to search for candidate matching cases of the current case; matching second keywords corresponding to each of the candidate matching cases with the knowledge graph to obtain second vector representations of each of the candidate matching cases; based on a preset similarity model, obtaining the similarity between the first vector representation and each of the second vector representations, and determining a set of matching cases from the candidate matching cases according to the similarity, so as to process the current case based on the set of matching cases.

[0008] In an exemplary embodiment of the present disclosure, extracting target keywords in the current case data, and matching the target keywords with a pre-constructed knowledge graph to obtain a first vector representation of the current case, includes: obtaining target entity nodes in the knowledge graph that have a mapping relationship with the target keywords, entity attributes corresponding to the target entity nodes, and inter-entity relationships of the target entity nodes, where entity nodes in the knowledge graph have corresponding vector representations; fusing the vector representations corresponding to the target entity nodes according to the entity attributes and the inter-entity relationships to obtain the first vector representation of the current case.

[0009] In an exemplary embodiment of the present disclosure, based on a preset similarity model, obtaining the similarity between the first vector representation and each of the second vector representations, and determining a set of matching cases from the candidate matching cases according to the similarity, includes: inputting the first vector representation and the second vector representations into the preset similarity model to obtain the similarity, where the preset similarity model is obtained by training a similarity model using pre-constructed vector representations of business cases; obtaining target second vector representations corresponding to similarities greater than a preset similarity threshold, and generating the set of matching cases according to the candidate matching cases corresponding to the target second vector representations.

[0010] In an exemplary embodiment of the present disclosure, after determining the set of matching cases from the candidate matching cases according to the similarity, the method further includes: obtaining the business types and corresponding business processing decisions of each matching case in the set of matching cases from the matching case data corresponding to the set of matching cases; pushing a target matching case in the set of matching cases to the user according to the business types and the business processing decisions.

[0011] In an exemplary embodiment of the present disclosure, pushing a target matching case in the set of matching cases to the user according to the business types and the business processing decisions, includes: determining whether there is a first target matching case in the set of matching cases that has a target business type and at the same time has a target business processing decision; if so, pushing the first target matching case to the user; if not, calculating the risk values of each matching case in the set of matching cases according to the business types and the business processing decisions of each matching case in the set of matching cases according to a preset risk assessment rule, and determining and pushing a second target matching case to the user according to the risk values.

[0012] In an exemplary embodiment of the present disclosure, the method of calculating the risk value of each matching case according to the preset risk assessment rule based on the business type and business processing decision of each matching case in the matching case set, and determining and pushing the second target matching case to the user according to the risk value includes: obtaining the type weight factor corresponding to the business type of each matching case and the decision weight factor corresponding to the business processing decision of each matching case according to the preset risk assessment rule; for each matching case, calculating the risk value of each matching case according to the corresponding type weight factor, decision weight factor, type weight factor and decision weight factor corresponding to the target business type; obtaining the second target matching cases with risk values less than the preset risk threshold and pushing them to the user; or obtaining the second target matching cases with risk values greater than the preset risk threshold and pushing them to the user.

[0013] In an exemplary embodiment of the present disclosure, the construction process of the pre-constructed knowledge graph includes: obtaining business case data, determining business entities in the business case data, and constructing entity attributes for each business entity; establishing mapping relationships between the business entities to obtain relationships between entities; determining at least one triple according to each business entity, the entity attributes corresponding to the business entity, and the relationships between entities; constructing the knowledge graph based on the at least one triple.

[0014] In an exemplary embodiment of the present disclosure, the construction process further includes: inputting the knowledge graph into a pre-trained graph convolutional neural network to obtain vector representations corresponding to business entity nodes in the knowledge graph.

[0015] According to one aspect of the present disclosure, there is provided a case processing device based on a knowledge graph, the device including: a first vector representation module, configured to extract target keywords in current case data, and match the target keywords with a pre-constructed knowledge graph to obtain a first vector representation of the current case; a case indexing module, configured to input the target keywords into a pre-constructed case index and search for candidate matching cases of the current case; a second vector representation module, configured to match second keywords corresponding to the candidate matching cases with the knowledge graph to obtain second vector representations of the candidate matching cases; a case processing module, configured to obtain the similarity between the first vector representation and each of the second vector representations based on a preset similarity model, and determine a matching case set from the candidate matching cases according to the similarity, so as to process the current case based on the matching case set.

[0016] According to one aspect of the present disclosure, there is provided a computer storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the above-mentioned case processing method based on a knowledge graph as described in any one of the above is implemented.

[0017] According to one aspect of the present disclosure, there is provided an electronic device, including: one or more processors; and a storage device for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the case processing method based on a knowledge graph as described in any one of the above.

[0018] In the case processing method based on a knowledge graph in an exemplary embodiment of the present disclosure, a target keyword in current case data is matched with a pre-constructed knowledge graph to obtain a first vector representation of the current case, and then the target keyword is input into a pre-constructed case index to search for candidate matching cases of the current case and obtain a second vector representation of the candidate matching cases. Finally, the similarity between the first vector representation and the second vector representation is calculated to determine a set of matching cases from the candidate matching cases according to the similarity and feed it back to the user for processing the current case.

[0019] On the one hand, based on a pre-constructed knowledge graph, cases are vectorized and a set of matching cases is determined based on the similarity with the vector representations of pre-stored existing cases, realizing semantic retrieval of cases and improving the accuracy of case retrieval; at the same time, the target keyword is input into a pre-constructed knowledge graph case index for preliminary screening to obtain candidate matching cases, and a set of matching cases is determined through a preset similarity model based on the candidate matching cases. The accuracy of obtaining the set of matching cases is improved through a two-step screening method; on the other hand, the obtained set of matching cases is automatically pushed to the user for processing the current case, avoiding the consumption of human resources caused by manual search, reducing the work burden of the user, and improving the business handling efficiency of the user.

[0020] It should be understood that the above general description and the following detailed description are only exemplary and explanatory and cannot limit the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present disclosure will become readily understood. In the drawings, several embodiments of the present disclosure are shown in an exemplary but not restrictive manner, wherein:

[0022] Figure 1 A flowchart of a case processing method based on a knowledge graph according to an exemplary embodiment of the present disclosure is shown;

[0023] Figure 2Shows a flowchart of constructing a knowledge graph according to an exemplary embodiment of the present disclosure;

[0024] Figure 3 Shows a flowchart of obtaining a first vector representation of the current case by matching a target keyword with a pre-constructed knowledge graph according to an exemplary embodiment of the present disclosure;

[0025] Figure 4 Shows a flowchart of constructing a similarity model according to an exemplary embodiment of the present disclosure;

[0026] Figure 5 Shows a flowchart of obtaining a target matching case according to an exemplary embodiment of the present disclosure;

[0027] Figure 6 Shows a flowchart of determining a target matching case according to a business type and a business processing decision and pushing it to a user according to an exemplary embodiment of the present disclosure;

[0028] Figure 7 Shows a flowchart of obtaining a second target matching case and pushing it to a user according to an exemplary embodiment of the present disclosure;

[0029] Figure 8 Shows a schematic structural diagram of a case processing device based on a knowledge graph according to an exemplary embodiment of the present disclosure;

[0030] Figure 9 Shows a schematic diagram of a storage medium according to an exemplary embodiment of the present disclosure; and

[0031] Figure 10 Shows a block diagram of an electronic device according to an exemplary embodiment of the present disclosure.

[0032] In the drawings, the same or corresponding reference numerals indicate the same or corresponding parts. Detailed Embodiment

[0033] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be more complete and comprehensive, and will fully convey the concept of the exemplary embodiments to those skilled in the art. Like reference numerals in the figures denote like or similar structures and thus their detailed description will be omitted.

[0034] In addition, the described features, structures, or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, numerous specific details are provided to give a thorough understanding of the embodiments of the present disclosure. However, those skilled in the art will realize that the technical solutions of the present disclosure may be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, well-known structures, methods, devices, implementations, or operations are not shown or described in detail to avoid obscuring aspects of the present disclosure.

[0035] The block diagrams shown in the drawings are only functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or these functional entities or a part of the functional entities may be implemented in one or more software-hardened modules, or these functional entities may be implemented in different networks and / or processor devices and / or microcontroller devices.

[0036] In many industries, such as banks, securities companies, insurance companies, trust investment companies, and fund management companies, etc., a lot of related or similar cases are often involved in business processing work, and these related or similar cases often have an auxiliary effect on users' processing of current business. Taking the insurance industry as an example, when receiving an insurance claim provided by a customer to the company, the claims adjuster can look up the case information or processing decisions of related cases as a reference for this insurance claim, that is, it can assist the claims adjuster to improve the processing efficiency of the current business. Currently, the process of finding related or similar cases includes: first, the claims adjuster uses keywords to retrieve cases, and then conducts manual verification. This not only takes time and effort, but also lacks semantic-level matching of cases only through keyword retrieval, and the accuracy of case finding is low, thus affecting the processing efficiency and effect of the business.

[0037] Based on this, in an exemplary embodiment of the present disclosure, first, a case processing method based on a knowledge graph is provided. Refer to Figure 1 As shown, the case processing method based on the knowledge graph includes the following steps:

[0038] Step S110: Extract the target keywords in the current case data, and match the target keywords with a pre-constructed knowledge graph to obtain the first vector representation of the current case;

[0039] Step S120: Input the target keywords into a pre-constructed case index, search to obtain the candidate matching cases of the current case and the corresponding second vector representations of the candidate matching cases, where the vector representations corresponding to the cases in the case index have been pre-stored;

[0040] Step S130: Based on a preset similarity model, obtain the similarities between the first vector representation and each second vector representation, and determine a set of matching cases from the candidate matching cases according to the similarities, so as to process the current case based on the set of matching cases.

[0041] According to the case processing method based on a knowledge graph in this exemplary embodiment, based on a pre-constructed knowledge graph, the cases are vectorized, and a set of matching cases is determined based on the similarities with the vector representations of the existing cases stored in advance, realizing semantic retrieval of cases and improving the accuracy of case retrieval. At the same time, by inputting the target keywords into the pre-constructed knowledge graph case index for preliminary screening first, candidate matching cases are obtained, and based on the candidate matching cases, a set of matching cases is determined through a preset similarity model. The accuracy of obtaining the set of matching cases is improved through a two-step screening method. The obtained set of matching cases is automatically pushed to the user for processing the current case, avoiding the consumption of human resources caused by manual search, reducing the work burden of the user, and improving the business handling efficiency of the user.

[0042] The following combines Figure 1 to illustrate the case processing method based on a knowledge graph in the exemplary embodiments of the present disclosure.

[0043] In step S110, extract the target keywords in the current case data, and match the target keywords with the pre-constructed knowledge graph to obtain the first vector representation of the current case.

[0044] In the exemplary embodiments of the present disclosure, the current case data is the description information of the current case; the target keywords are words that are distinguishable and exist independently in the current case. The target keywords in the current case data are extracted through a keyword extraction algorithm, and the keyword extraction algorithm includes but is not limited to the TF-IDF (term frequency-inverse document frequency) algorithm, the keyword extraction based on the pyhanlp algorithm (a word segmentation tool); or, the target keywords are determined by extracting the entities in the current case data. Entity extraction is named entity recognition, including but not limited to CRF (Conditional Random Fields), CNN + CRF (Convolutional Neural Network + Conditional Random Fields), BiLSTM + CRF (Bidirectional Long Short-Term Memory Model + Conditional Random Fields), etc. The present disclosure does not make special limitations on the method of extracting keywords.

[0045] The knowledge graph is a structured semantic knowledge base used to quickly describe case concepts (entities) and their interrelationships. That is, the knowledge graph is a network graph structure composed of many concept (entity) nodes and edges. The edges in the graph identify the relationships between the two connected nodes. The knowledge graph based on triples is a common representation method. The basic forms of triples include (entity-entity relationship-entity) and (entity-attribute-attribute value), etc. Among them, the concept (entity) is further described by defining attributes for the concept (entity). The entity relationship connects different entities, and the nodes in the knowledge graph are connected by entity relationship nodes to form a graph. For example, the capital of China is Beijing, where China and Beijing are entities, and capital is the relationship between the entities China and Beijing.

[0046] In an exemplary embodiment of the present disclosure, the construction process of the pre-constructed knowledge graph includes: First, obtain business case data, determine business entities in the business case data, and construct entity attributes for each business entity; Second, establish mapping relationships between the business entities to obtain inter-entity relationships; Then, determine at least one triple according to the business entities, the entity attributes corresponding to the business entities, and the inter-entity relationships; Finally, construct a knowledge graph based on at least one triple. Taking the insurance industry as an example, business entities include insurance policies, claims cases, personnel, institutions, etc.; entity attributes such as the processing status of claim reports, the names, ages, and ID numbers of policyholders, etc., are information used to further refine the description of business entities; inter-entity relationships such as policyholders of insurance policies, beneficiaries, and claim reports of insurance policies. The formed triples are, for example, (insurance policy-beneficiary-person) or (person-age-29), etc. Then, based on the obtained triples, a knowledge graph is constructed.

[0047] In some possible implementation manners, a mapping relationship can be established between the business entity nodes, entity attribute nodes, and inter-entity relationship nodes in the knowledge graph and database tables. Based on this, entities (concepts) can be extracted from the database tables at a set period to construct and update the knowledge graph according to the database tables at a set period. That is to say, structured case data can be obtained from the database of business processing, and the business entities corresponding to the structured case data can be determined.

[0048] In some possible implementation manners, in order to make the constructed knowledge graph large enough, in addition to obtaining structured case data from the business processing database, relevant data of business processing can also be obtained from the network or other platforms, etc. Specifically, relevant data of business processing in web pages can be obtained to obtain unstructured case data, and the unstructured case data can be processed based on natural language processing algorithms for constructing the knowledge graph. It should be noted that there may be duplicate business entities in the obtained structured business entities and unstructured business entities, and the duplicate business entities can also be de-duplicated.

[0049] In an exemplary embodiment of the present disclosure, after constructing a knowledge graph according to at least one triple, the knowledge graph can also be input into a pre-trained graph convolutional neural network to obtain a vector representation corresponding to the business entity nodes in the knowledge graph.

[0050] Specifically, first, based on the pre-constructed knowledge graph, the number of business entity nodes in the knowledge graph is N, and each business entity node has a corresponding entity attribute D. Then, the business entity nodes in the knowledge graph and the corresponding entity attributes form an N×D-dimensional matrix V. Secondly, an N×N-dimensional matrix E, also called an adjacency matrix, is formed according to the inter-entity relationships between the business entity nodes. Then, the knowledge graph can be represented as G(V, E). Finally, the knowledge graph G(V, E) is input into the graph convolutional neural network, that is, the matrices V and E are used as the inputs of the graph convolutional neural network for training. Among them, during the training process, for each business entity node, feature information (including the features of the business entity node itself) is obtained from the neighbor nodes of the business entity node, so as to return a result vector as the vector representation of the business entity node by using the features of the neighbor nodes. Among them, in each iteration process, the training error is gradually reduced based on an unsupervised loss function, and the vector weights corresponding to the business entity nodes are updated according to the entity attributes corresponding to the business entity nodes. Continuous training is performed until the number of iterations reaches a preset iteration threshold, and then the vector representations of each business entity node in the knowledge graph are output.

[0051] Further, after obtaining the vector representations of each business entity node in the knowledge graph, a business case vector representation is constructed for the existing cases. Specifically, based on the vector representations of each business entity node in the knowledge graph, the attributes of the concepts (entities) and the inter-entity relationships are extracted from the case data of the existing cases, and the vector representations corresponding to the concept (entity) nodes extracted from the existing cases are fused to obtain the business case vector representation corresponding to the existing cases. Further, after generating the corresponding business case vector representation for each existing case, the vector representations of each existing case are stored in the knowledge graph case index that has been constructed. Thus, the construction of the knowledge graph is completed. That is to say, referring to the knowledge graph construction flow chart as shown in Figure 2 The construction process of the knowledge graph includes: the process of obtaining business case data, extracting concepts (entities) and establishing inter-entity relationships in the business case data, constructing the knowledge graph, representing business case vectors, and establishing the knowledge graph case index.

[0052] Through the exemplary embodiments of the present disclosure, a knowledge graph based on processing business is established, a case index is established for the graph, and the business case vectors of the existing cases are represented and stored in the index, so as to realize the construction of the knowledge graph for subsequent case processing based on the knowledge graph.

[0053] Further, in an exemplary embodiment of the present disclosure, after extracting the target keyword in the current case data, it is matched with a pre-constructed knowledge graph to obtain a first vector representation of the current case. Figure 3 The flowchart shows the process of obtaining the first vector representation of the current case by matching the target keyword with a pre-constructed knowledge graph according to an exemplary embodiment of the present disclosure, as Figure 3 shown, and this process includes the following steps:

[0054] In step S310, obtain the target entity node in the knowledge graph that has a mapping relationship with the target keyword, the entity attributes corresponding to the target entity node, and the inter-entity relationships of the target entity node.

[0055] In an exemplary embodiment of the present disclosure, it can be known from the above process of constructing the knowledge graph that the entity nodes in the knowledge graph have corresponding vector representations. Therefore, the obtained target keyword can be used to match with the knowledge graph to obtain the target entity node with a mapping relationship. This mapping relationship is the same relationship as the target keyword or has a semantic similarity relationship with the target keyword. For example, "claims settlement" and "compensation" have a semantic similarity relationship. After obtaining the target entity node with a mapping relationship with the target keyword, correspondingly, obtain the entity attributes corresponding to the target entity node and the inter-entity relationships it has.

[0056] In step S320, according to the entity attributes and inter-entity relationships, fuse the vector representations corresponding to the target entity nodes to obtain the first vector representation of the current case.

[0057] In an exemplary embodiment of the present disclosure, fuse the vector representations of the target entity nodes according to the entity attributes and inter-entity relationships corresponding to the target entity to obtain the first vector representation of the current case. Specifically, the target entity nodes can be combined according to the entity attributes and inter-entity relationships to obtain the first vector representation of the current case.

[0058] According to this exemplary embodiment, based on the extracted target keyword of the current case and based on the pre-constructed knowledge graph, a first vector representation containing semantics corresponding to the current case can be obtained, thereby improving the accuracy of subsequent case queries based on this first vector representation.

[0059] In step S120, input the target keyword into the pre-constructed knowledge graph case index, and search to obtain the candidate matching cases of the current case and the corresponding second vector representations of the candidate matching cases.

[0060] In an exemplary embodiment of the present disclosure, as can be seen from the above process of building the knowledge graph, the cases in the constructed case index have pre-stored corresponding vector representations. Therefore, after obtaining the candidate matching cases of the current case based on the target keyword search, correspondingly, the second vector representations corresponding to the candidate matching cases are obtained. Among them, the constructed case index is an inverted index built based on existing cases. The inverted index is a specific storage form for implementing the "keyword-document matrix". Through the inverted index, a document list containing the keyword can be quickly obtained according to the keyword, that is, the mapping from the keyword to the file id (identity document).

[0061] In step S130, based on a preset similarity model, the similarities between the first vector representation and each second vector representation are obtained, and a set of matching cases is determined from the candidate matching cases according to the similarities, so as to process the current case based on the set of matching cases.

[0062] In an exemplary embodiment of the present disclosure, the preset similarity model is obtained by training a similarity model using pre-constructed business case vector representations. First, the first vector representation and the second vector representation are input into the preset similarity model to obtain the similarity between the first vector representation and the second vector representation. Secondly, the target second vector representations corresponding to the similarities greater than the preset similarity threshold are obtained, and a set of matching cases is generated according to the candidate matching cases corresponding to the target second vector representations.

[0063] The following combines Figure 4 to illustrate the construction process of the similarity model in the exemplary embodiment of the present disclosure. Among them, the cosine similarity model is taken as an example for the similarity model for illustration. As Figure 4 shown, the construction process of the similarity model includes:

[0064] In step S410, information annotation is performed on existing cases, and corresponding business case vector representations are constructed to obtain a dataset of similar cases; in step S420, the dataset of similar cases is divided into a training set, a validation set, and a test set, and the division ratio can be 70%-15%-15%, 80%-10%-10%, etc. The present disclosure does not make special limitations on this; in step S430, hyperparameters and other training parameters of the similarity model are set; in step S440, the similarity model is trained and evaluated, and the dataset of similar cases is adjusted and the hyperparameters are adjusted according to the training results until the model with the best performance on the validation set is obtained as the trained preset similarity model.

[0065] In an exemplary embodiment of the present disclosure, after determining the set of matching cases from the candidate matching cases according to the similarities, a target matching case can also be obtained from the candidate matching cases and pushed to the user, so that the user can process the current case according to the target matching case.

[0066] Figure 5 shows a flowchart of obtaining a target matching case according to an exemplary embodiment of the present disclosure, as Figure 5 shown, the process includes:

[0067] In step S510, from the matching case data corresponding to the matching case set, obtain the business types of each matching case in the matching case set and the corresponding business processing decisions.

[0068] In an exemplary embodiment of the present disclosure, after obtaining the matching case set, obtain the business type of each matching case in the matching case set and the corresponding business processing decision; wherein, the business type is used to identify the business category to which the matching case belongs. Taking the insurance industry as an example, the business types may include accident insurance, health insurance, life insurance, etc., and the business processing decision is used to identify the processing result of the corresponding matching case. Taking the insurance industry as an example again, the business processing decisions may include claim settlement, claim settlement amounts at different levels, rejection of claim settlement, etc.

[0069] Through this exemplary embodiment, since there may be a large number of matching cases in the obtained matching case set, and the business types and business processing decisions of each matching case may be different, by obtaining the business type and business processing decision of each matching case in the matching case set, the information of each matching case is controlled from the business dimension for case analysis.

[0070] In step S520, according to the business type and business processing decision, push the target matching cases in the matching case set to the user.

[0071] In an exemplary embodiment of the present disclosure, the target matching cases can be determined from the matching case set according to the business type and business processing decision of each matching case and pushed to the user.

[0072] Figure 6 shows a flowchart of determining target matching cases according to the business type and business processing decision and pushing them to the user according to an exemplary embodiment of the present disclosure, as Figure 6 shown, the process includes:

[0073] In step S610, determine whether there is a first target matching case in the matching case set that has the target business type and at the same time has the target business processing decision.

[0074] In an exemplary embodiment of the present disclosure, the target business type can be determined according to the actual situation of processing the business. Taking the processing of insurance claims as an example, the target business type can be the business type of the current case where the customer submits a claim request. If the customer submits a claim for accidental injury insurance, the target business type can be accidental injury insurance. The target business processing decision can be determined according to the result requirements of business processing. Still taking the processing of insurance claims as an example, for claims handlers, being able to quickly obtain cases with the same business type as the current claim case and with a business processing decision of rejecting the claim can provide a reference for processing the current case. Therefore, for processing insurance claims, the target business processing decision can be claim settlement, rejection of claim, claim amount, etc.

[0075] In step S620, if there is a first target matching case in the matching case set that has the target business type and simultaneously has the target business processing decision, the first target matching case is pushed to the user.

[0076] In an exemplary embodiment of the present disclosure, the first target matching case is a matching case that has the target business type and simultaneously has the target business processing decision. The first target matching case is pushed to the user so that the user can use the first target matching case as a reference to process the current business, thereby improving the processing efficiency of the user for the current business.

[0077] In step S630, if there is no first target matching case in the matching case set that has the target business type and simultaneously has the target business processing decision, then according to the preset risk assessment rules, based on the business type and business processing decision of each matching case in the matching case set, calculate the risk value of each matching case, and determine and push a second target matching case to the user according to the risk value.

[0078] In an exemplary embodiment of the present disclosure, the risk value of a matching case is used to identify the deviation degree between the matching case and the target processing decision of the current case. The higher the risk value of the matching case, the greater the deviation degree between the matching case and the target processing decision of the current case. On the contrary, the lower the risk value of the matching case, the smaller the deviation degree from the target processing decision of the current case. The type weight factor can be used to identify the similarity degree of the business types of different matching cases. That is to say, the closer the type weight factors are, the higher the similarity degree of the business types of the corresponding matching cases. And if the type weight factors are the same, it indicates that the business types of the corresponding matching cases are the same.

[0079] In some possible implementation manners, according to the actual business processing requirements, the higher the risk value of a matching case, the more suitable the matching case with high risk is as a reference case for the current case.

[0080] In some possible implementation manners, according to the actual processing service requirements, the lower the risk value of the matching case, the more suitable the matching case with low risk is as a reference case for the current case.

[0081] In an exemplary embodiment of the present disclosure, Figure 7 A flowchart of obtaining a second target matching case and pushing it to a user according to an exemplary embodiment of the present disclosure is shown, as Figure 7 shown, and the process includes:

[0082] In step S710, according to a preset risk assessment rule, obtain a type weight factor corresponding to the service type to which each matching case belongs, and a decision weight factor corresponding to the service processing decision of each matching case.

[0083] In an exemplary embodiment of the present disclosure, the preset risk assessment factor is preconfigured according to service processing requirements and stipulates the type weight factor corresponding to various service types and the decision weight factor corresponding to the service processing decision. Based on this, according to the preset risk assessment rule, the type weight factor corresponding to the service type to which each matching case belongs, and the decision weight factor corresponding to the service processing decision of each matching case can be obtained.

[0084] In step S720, for each matching case, calculate the risk value of each matching case according to the corresponding type weight factor, decision weight factor, type weight factor corresponding to the target service type, and decision weight factor.

[0085] In an exemplary embodiment of the present disclosure, for each matching case, the risk value of each matching case can be calculated according to the corresponding type weight factor, decision weight factor, type weight factor corresponding to the target service type, and decision weight factor.

[0086] In some possible implementation manners, respectively obtain a first difference value between each matching case and the type weight factor corresponding to the target service type, and obtain a second difference value between each matching case and the decision weight factor corresponding to the target service processing decision; then obtain the weighted sum of the first difference value and the second difference value, where the weights of the first difference value and the second difference value are respectively determined by the preset service type weight and the weight of the service processing decision, and the weight tendency of the service type and the service processing decision can be adjusted according to actual service requirements, so as to realize the configurable tendency of the risk value.

[0087] Optionally, the first difference value can be respectively the absolute value of the difference between each matching case and the type weight factor corresponding to the target service type, and the second difference value can be respectively the absolute value of the difference between each matching case and the decision weight factor corresponding to the target service type.

[0088] Optionally, the first gap value can be obtained by taking the absolute value of the difference between each matching case and the type weight factor corresponding to the target business type, and then dividing it by the type weight factor corresponding to the target business type. The second gap value can be obtained by taking the absolute value of the difference between each matching case and the decision weight factor corresponding to the target business type, and then dividing it by the decision weight factor corresponding to the target business type. Of course, the first gap value and the second gap value can also be calculated in other ways. The present disclosure includes, but is not limited to, the above calculation methods for the first gap value and the second gap value.

[0089] Taking the processing of insurance claims as an example, the risk value of each matching case is calculated according to the corresponding type weight factor, decision weight factor, type weight factor corresponding to the target business type, and decision weight factor corresponding to the target business type. For example, if a customer A has an accident insurance policy and fractures their wrist while skating, incurring certain medical expenses. The skating rink ticket includes accident insurance, and the medical expenses are within the insurance limit, so the skating rink pays the customer's medical expenses. However, the customer further claims for accident insurance from the insurance company.

[0090] First, determine that the target business type of this claim business is "accident insurance", and according to the business processing requirements, the target business processing decision is "reject the claim". According to the preset risk assessment rules, the target type weight factor corresponding to the target business type "accident insurance" is 7, and the target decision weight factor corresponding to the target business processing decision "reject the claim" is 10. Also, the tendency weights of the business type and the business processing decision are 70% and 30% respectively.

[0091] Secondly, after determining the set of matching cases based on the current claim case data, it is determined that there is no first target matching case with "accident insurance" and "reject the claim" in the set of matching cases. Then, according to the preset risk assessment rules, the weight factors corresponding to the business types of each matching case in the set of matching cases are obtained as follows: the type weight factor of matching case 1 is 6, the type weight factor of matching case 2 is 4, and the type weight factor of matching case 3 is 2. Correspondingly, the decision weight factors corresponding to each matching case in the set of matching cases are obtained as follows: the decision weight factor of matching case 1 is 6, the decision weight factor of matching case 2 is 10, and the decision weight factor of matching case 3 is 8. Then, the risk values of matching case 1, matching case 2, and matching case 3 are calculated as follows: 1.9 = |7 - 6|×70% + |10 - 6|×30%, 2.1 = |7 - 4|×70% + |10 - 10|×30%, and 4.1 = |7 - 2|×70% + |10 - 8|×30%. Of course, the type weight factors and decision weight factors in this exemplary embodiment are only exemplary and can be adjusted according to actual business processing requirements.

[0092] Through this exemplary embodiment, the risk value of each matching case can be calculated according to the type weight factor and decision weight factor of each matching case. This calculation process combines the business type of each matching case and the corresponding business processing decision, and the difference value from the target business type and the corresponding target business processing decision, so as to consider the risk value of each matching case from the dimensions of case business type and business processing decision, so as to subsequently determine the second target matching case from the set of matching cases according to the risk value, and further improve the similarity between the second target matching case and the current case from the dimensions of case business type and business processing decision.

[0093] In step S730, the second target matching cases with risk values less than the preset risk threshold or greater than the preset risk threshold are obtained and pushed to the user.

[0094] In the exemplary embodiment of the present disclosure, the second target matching cases with risk values less than the preset risk threshold can be obtained and pushed to the user; alternatively, the second target matching cases with risk values greater than the preset risk threshold can be obtained and pushed to the user.

[0095] It should be noted that which discrimination method is selected to determine the second target matching case to be pushed to the user can be determined according to the specific calculation method of the risk value and the actual business processing requirements.

[0096] Taking the above handling of insurance claims as an example, when the risk values of matching case 1, matching case 2, and matching case 3 are 1.9, 2.1, and 4.1 respectively, and the preset risk threshold is 3, the second target matching cases with risk values less than the preset risk threshold are matching case 1 and matching case 2, and matching case 1 and matching case 2 are pushed to the user. Among them, for a matching case with a risk value less than the preset risk threshold, it indicates that the corresponding matching case is more consistent with the current case in terms of business type and business processing decision. For example, the type weight factor of matching case 1 is 6, indicating that matching case 1 has a very high similarity with the current case (type weight factor is 7) in terms of case type. At the same time, the decision weight factor of matching case 1 is 6, indicating that there may be a certain gap between the business processing decision of matching case 1 and "rejecting the claim"; correspondingly, the type weight factor of matching case 2 is 4, indicating that matching case 1 has a certain similarity with the current case (type weight factor is 7) in terms of case type, and the decision weight factor of matching case 2 is 10, indicating that the business processing decision of matching case 2 is the same as the business processing decision of "rejecting the claim". The type weight factor of matching case 3 is 2, and the decision weight factor of matching case 3 is 8. Since the type weight factor of matching case 3 differs greatly from the target type weight factor 7, and the tendency weight corresponding to the business type is 70%, that is, the business type accounts for a relatively large proportion in the calculation of the risk value. Therefore, even though the decision weight factor of matching case 3 is 8, which is not much different from the target decision weight factor 10, the final risk value of matching case 3 obtained is 4.1, which is greater than the preset risk threshold, that is, matching case 3 is not pushed to the user as a second target matching case; on the contrary, since the type weight factors of matching case 1 and matching case 2 are close to the current case, matching case 1 and matching case 2 are pushed to the user.

[0097] According to this exemplary embodiment, the second target matching cases with risk values less than or greater than the preset risk threshold are the cases that are most consistent with the current case in terms of case type and business processing decision. By recommending the second target matching cases to the user, it can assist the user in handling the current case, thereby improving the handling effect of the current case. To a certain extent, in the case of having similar case references, it can also improve the persuasiveness of the handling result of the current case.

[0098] According to the case processing method based on the knowledge graph in the present exemplary embodiment, based on the pre-constructed knowledge graph, the case is vectorized and represented, and a set of matching cases is determined based on the similarity with the vector representation of the cases pre-stored, realizing the semantic retrieval of cases and improving the accuracy of case retrieval; at the same time, the target keywords are input into the pre-constructed case index of the knowledge graph for preliminary screening to obtain candidate matching cases, and on the basis of the candidate matching cases, a set of matching cases is determined through a preset similarity model. The accuracy of obtaining the set of matching cases is improved through a two-step screening method; the obtained set of matching cases is automatically pushed to the user to process the current case, avoiding the consumption of human resources caused by manual search, reducing the user's workload, and improving the user's business handling efficiency.

[0099] In addition, in an exemplary embodiment of the present disclosure, a case processing device based on a knowledge graph is also provided. Refer to Figure 8 As shown, the case processing device 800 based on the knowledge graph may include a vector representation module 810, a case index module 820, and a case processing module 830. Specifically,

[0100] The vector representation module 810 is configured to extract target keywords from the current case data, match the target keywords with the pre-constructed knowledge graph, and obtain a first vector representation of the current case;

[0101] The case index module 820 is configured to input the target keywords into the pre-constructed case index, search for candidate matching cases of the current case and the corresponding second vector representations of the candidate matching cases, where the cases in the case index have pre-stored corresponding vector representations;

[0102] The case processing module 830 is configured to obtain the similarity between the first vector representation and each of the second vector representations based on a preset similarity model, and determine a set of matching cases from the candidate matching cases according to the similarity, so as to process the current case based on the set of matching cases.

[0103] In an exemplary embodiment of the present disclosure, the vector representation module 810 may include: a node acquisition unit configured to acquire target entity nodes in the knowledge graph that have a mapping relationship with the target keywords, entity attributes corresponding to the target entity nodes, and inter-entity relationships of the target entity nodes, where the entity nodes in the knowledge graph have corresponding vector representations; a vector fusion unit configured to fuse the vector representations corresponding to the target entity nodes according to the entity attributes and the inter-entity relationships to obtain a first vector representation of the current case.

[0104] In an exemplary embodiment of the present disclosure, the case processing module 830 may include: a similarity acquisition unit configured to input the first vector representation and the second vector representation into the preset similarity model to obtain the similarity, where the preset similarity model is obtained by training a similarity model using pre-constructed business case vector representations; a comparison unit configured to obtain the target second vector representation corresponding to the similarity greater than the preset similarity threshold, and generate the matching case set according to the candidate matching cases corresponding to the target second vector representation.

[0105] In an exemplary embodiment of the present disclosure, the case processing apparatus 800 based on the knowledge graph may further include: an acquisition module configured to acquire the business types and corresponding business processing decisions of each matching case in the matching case set from the matching case data corresponding to the matching case set; a recommendation module configured to push the target matching case in the matching case set to the user according to the business type and the business processing decision.

[0106] In an exemplary embodiment of the present disclosure, the recommendation module may include: a judgment unit configured to judge whether there is a first target matching case in the matching case set that has the target business type and at the same time has the target business processing decision; a first push unit configured to, if there is a first target matching case that has the target business type and at the same time has the target business processing decision, push the first target matching case to the user; a second push unit configured to, if there is no first target matching case, calculate the risk value of each matching case in the matching case set according to the business type and the business processing decision of each matching case in the matching case set according to the preset risk assessment rule, and determine and push a second target matching case to the user according to the risk value.

[0107] In an exemplary embodiment of the present disclosure, the recommendation module may further include: a weight factor acquisition unit configured to acquire the type weight factor corresponding to the business type of each matching case and the decision weight factor corresponding to the business processing decision of each matching case according to the preset risk assessment rule; a risk value calculation unit configured to calculate the risk value of each matching case according to the corresponding type weight factor, decision weight factor, the type weight factor and decision weight factor corresponding to the target business type for each matching case; a second target matching case acquisition unit configured to acquire the second target matching case with a risk value less than the preset risk threshold and push it to the user; or acquire the second target matching case with a risk value greater than the preset risk threshold and push it to the user.

[0108] In an exemplary embodiment of the present disclosure, the case processing apparatus 800 based on a knowledge graph may further include: an information extraction module, configured to obtain business case data, determine business entities in the business case data, and construct entity attributes for each of the business entities; a relationship establishment module, configured to establish mapping relationships between the business entities to obtain relationships between entities; the relationship establishment module is further configured to determine at least one triple according to each of the business entities, the entity attributes corresponding to the business entities, and the relationships between entities; a graph construction module, configured to construct the knowledge graph based on the at least one triple.

[0109] In an exemplary embodiment of the present disclosure, the case processing apparatus 800 based on a knowledge graph may further include: an entity node vector representation module, configured to input the knowledge graph into a pre-trained graph convolutional neural network to obtain a vector representation corresponding to a business entity node in the knowledge graph.

[0110] Since each functional module of the case processing apparatus based on a knowledge graph in the exemplary embodiment of the present disclosure is the same as that in the inventive embodiment of the above-mentioned case processing method based on a knowledge graph, details thereof will not be described herein again.

[0111] It should be noted that although several modules or units of the case processing apparatus based on a knowledge graph are mentioned in the above detailed description, such a division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of the two or more modules or units described above may be embodied in one module or unit. Conversely, the features and functions of one module or unit described above may be further divided and embodied by multiple modules or units.

[0112] In addition, in an exemplary embodiment of the present disclosure, a computer storage medium capable of implementing the above method is further provided. A program product capable of implementing the above method of this specification is stored thereon. In some possible embodiments, various aspects of the present disclosure may also be implemented in the form of a program product, which includes program code. When the program product runs on a terminal device, the program code is used to cause the terminal device to execute the steps according to various exemplary embodiments of the present disclosure described in the above "exemplary method" section of this specification.

[0113] Reference Figure 8 As shown, a program product 800 for implementing the above method according to an exemplary embodiment of the present disclosure is described. It may be a portable compact disc read-only memory (CD-ROM) and includes program code, and may run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto. In this document, a readable storage medium may be any tangible medium that contains or stores a program, and the program may be used by or in combination with an instruction execution system, apparatus, or device.

[0114] The program product may employ any combination of one or more readable media. The readable media may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0115] The computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which the readable program code is carried. Such a propagated data signal may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. The readable signal medium may also be any readable medium other than the readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device.

[0116] The program code contained on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0117] The program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and also including conventional procedural programming languages such as the "C" language or similar programming languages. The program code may execute entirely on the user computing device, partly on the user device, as a stand-alone software package, partly on the user computing device and partly on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0118] In addition, in an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided. Those skilled in the art can understand that various aspects of the present disclosure can be implemented as a system, a method, or a program product. Therefore, various aspects of the present disclosure can be specifically implemented in the following forms, namely: a complete hardware embodiment, a complete software embodiment (including firmware, microcode, etc.), or an embodiment combining hardware and software aspects, which can be collectively referred to as "circuitry", "module", or "system" here.

[0119] Reference will now be made to Figure 10 describe the electronic device 1000 according to such an embodiment of the present disclosure. Figure 10 The electronic device 1000 shown is merely an example and should not impose any limitation on the functions and usage scope of the embodiments of the present disclosure.

[0120] As Figure 10 shown, the electronic device 1000 is presented in the form of a general-purpose computing device. The components of the electronic device 1000 may include, but are not limited to: at least one of the above-mentioned processing units 1010, at least one of the above-mentioned storage units 1020, a bus 1030 connecting different system components (including the storage unit 1020 and the processing unit 1010), and a display unit 1040.

[0121] Among them, the storage unit stores program code, and the program code can be executed by the processing unit 1010, so that the processing unit 1010 executes the steps according to various exemplary embodiments of the present disclosure described in the above "Exemplary Method" section of this specification.

[0122] The storage unit 1020 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 1021 and / or a cache storage unit 1022, and may further include a read-only storage unit (ROM) 1023.

[0123] The storage unit 1020 may further include a program / utility 1024 having a set (at least one) of program modules 1025. Such program modules 1025 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data. The implementation of a network environment may be included in each or some combination of these examples.

[0124] The bus 1030 may represent one or more of several types of bus structures, including a storage unit bus or a storage unit controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0125] The electronic device 1000 can also communicate with one or more external devices 1100 (such as a keyboard, a pointing device, a Bluetooth device, etc.), and can also communicate with one or more devices that enable a user to interact with the electronic device 1000, and / or communicate with any device that enables the electronic device 1000 to communicate with one or more other computing devices (such as a router, a modem, etc.). Such communication can be carried out through the input / output (I / O) interface 1050. Moreover, the electronic device 1000 can also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through the network adapter 1060. As shown in the figure, the network adapter 1060 communicates with other modules of the electronic device 1000 through the bus 1030. It should be understood that although not shown in the figure, other hardware and / or software modules can be used in combination with the electronic device 1000, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0126] Through the description of the above embodiments, those skilled in the art can easily understand that the exemplary embodiments described herein can be implemented by software, or can be implemented by the way of software combined with necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure can be embodied in the form of a software product, and the software product can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

[0127] In addition, the above-mentioned drawings are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, rather than for limiting purposes. It is easy to understand that the processes shown in the above-mentioned drawings do not indicate or limit the time sequence of these processes. Additionally, it is also easy to understand that these processes can be executed synchronously or asynchronously in, for example, multiple modules.

[0128] After considering the specification and practicing the invention disclosed herein, those skilled in the art will readily think of other embodiments of the present disclosure. The present disclosure aims to cover any variations, uses, or adaptations of the present disclosure, which follow the general principles of the present disclosure and include the well-known common knowledge or conventional technical means in the technical field not disclosed in the present disclosure. The specification and the embodiments are only regarded as exemplary, and the true scope and spirit of the present disclosure are pointed out by the claims.

[0129] It should be understood that the present disclosure is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present disclosure is only limited by the appended claims.

Claims

1. A case processing method based on a knowledge graph, characterized in that Including: Extract the target keywords from the current case data, match the target keywords with a pre-constructed knowledge graph, and obtain the first vector representation of the current case; Input the target keywords into the pre-constructed knowledge graph case index, search for the candidate matching cases of the current case and the corresponding second vector representations of the candidate matching cases, where the cases in the case index have pre-stored corresponding vector representations; Based on a preset similarity model, obtain the similarities between the first vector representation and each of the second vector representations, and determine a set of matching cases from the candidate matching cases according to the similarities, so as to process the current case based on the set of matching cases; Among them, the processing of the current case based on the set of matching cases includes: If there is a first target matching case with a target business type and a target business processing decision in the set of matching cases, push the first target matching case to the user; Otherwise, according to a preset risk assessment rule, obtain the type weight factors corresponding to the business types of each of the matching cases and the decision weight factors corresponding to the business processing decisions of each of the matching cases; the magnitude relationship of the type weight factors is used to identify the similarity degree of the business types of different matching cases; the magnitude relationship of the decision weight factors is used to identify the similarity degree of the business processing decisions of different matching cases; For each of the matching cases, calculate the risk value of each of the matching cases according to the corresponding type weight factor, decision weight factor, the type weight factor and decision weight factor corresponding to the target business type, and determine and push a second target matching case to the user according to the risk value; the risk value of the matching case is used to identify the deviation degree between the matching case and the target processing decision of the current case; Among them, calculating the risk value of each of the matching cases includes: Respectively obtain the first difference value between each of the matching cases and the type weight factor corresponding to the target business type, and obtain the second difference value between each of the matching cases and the decision weight factor corresponding to the target business processing decision; Obtain the weighted sum of the first difference value and the second difference value to obtain the risk value of each of the matching cases.

2. The method according to claim 1, wherein The extracting the target keywords from the current case data, matching the target keywords with a pre-constructed knowledge graph, and obtaining the first vector representation of the current case includes: Obtain the target entity nodes in the knowledge graph that have a mapping relationship with the target keywords, the entity attributes corresponding to the target entity nodes, and the inter-entity relationships of the target entity nodes, where the entity nodes in the knowledge graph have corresponding vector representations; According to the entity attributes and the inter-entity relationships, fuse the vector representations corresponding to the target entity nodes to obtain the first vector representation of the current case.

3. The method according to claim 1, wherein The obtaining the similarities between the first vector representation and each of the second vector representations based on a preset similarity model, and determining a set of matching cases from the candidate matching cases includes: Input the first vector representation and the second vector representation into the preset similarity model to obtain the similarity, where the preset similarity model is obtained by training a similarity model using pre-constructed business case vector representations; Obtain the target second vector representation corresponding to the similarity greater than the preset similarity threshold, and generate the matching case set according to the candidate matching cases corresponding to the target second vector representation.

4. The method according to claim 1, characterized in that, Before processing the current case based on the matching case set, the method further includes: From the matching case data corresponding to the matching case set, obtain the business types and corresponding business processing decisions of each matching case in the matching case set.

5. The method according to claim 1, characterized in that, The determining and pushing the second target matching case to the user according to the risk value includes: Obtain the second target matching case with the risk value less than the preset risk threshold and push it to the user; or obtain the second target matching case with the risk value greater than the preset risk threshold and push it to the user.

6. The method according to any one of claims 1 to 5, characterized in that The construction process of the pre-constructed knowledge graph includes: Obtain business case data, determine the business entities in the business case data, and construct entity attributes for each of the business entities; Establish mapping relationships between the business entities to obtain relationships between entities; Determine at least one triple according to each of the business entities, the entity attributes corresponding to the business entities, and the relationships between entities; Construct the knowledge graph based on the at least one triple.

7. The method according to claim 6, characterized in that The construction process further includes: Input the knowledge graph into a pre-trained graph convolutional neural network to obtain the vector representations corresponding to the business entity nodes in the knowledge graph.

8. A case processing device based on a knowledge graph, characterized in that, It includes: A vector representation module, configured to extract target keywords from the current case data, and match the target keywords with a pre-constructed knowledge graph to obtain the first vector representation of the current case; A case indexing module, configured to input the target keywords into a pre-constructed case index, search for the candidate matching cases of the current case and the corresponding second vector representations of the candidate matching cases, where the cases in the case index have pre-stored corresponding vector representations; A case processing module, configured to obtain the similarities between the first vector representation and each of the second vector representations based on a preset similarity model, and determine a matching case set from the candidate matching cases according to the similarities, so as to process the current case based on the matching case set; The case processing module is configured to execute: If there is a first target matching case in the matching case set that has the target business type and the target business processing decision, then push the first target matching case to the user; Otherwise, according to the preset risk assessment rules, obtain the type weight factors corresponding to the business types of each of the matching cases, and the decision weight factors corresponding to the business processing decisions of each of the matching cases; The magnitude relationship of the type weight factors is used to identify the similarity degree of the business types of different matching cases; The magnitude relationship of the decision weight factors is used to identify the similarity degree of the business processing decisions of different matching cases; For each of the said matching cases, calculate the risk value of each matching case according to the corresponding type weight factor, decision weight factor, the type weight factor and decision weight factor corresponding to the target business type, and determine and push the second target matching case to the user according to the risk value; The risk value of the matching case is used to identify the deviation degree between the matching case and the target processing decision of the current case; Among them, calculating the risk value of each matching case includes: Obtain the first difference value between each matching case and the type weight factor corresponding to the target business type respectively, and obtain the second difference value between each matching case and the decision weight factor corresponding to the target business processing decision; Obtain the weighted sum of the first difference value and the second difference value to obtain the risk value of each matching case.

9. An electronic device, characterized in that, Including: One or more processors; A storage device for storing one or more programs, which when executed by the one or more processors, cause the one or more processors to implement the case processing method based on a knowledge graph as described in any one of claims 1 to 7.

Images