Knowledge graph construction method, application method and corresponding device

Abstract

The present disclosure relates to a method for constructing a knowledge graph, a method for applying the knowledge graph, and corresponding devices, to solve the problem of low efficiency and reliability of labor arbitration due to different experience levels of arbitrators in related technologies. The method for constructing a knowledge graph comprises: obtaining labor arbitration case data; determining a target element entity, a target clause entity, and a target reasoning template entity according to a preset target request type entity and the labor arbitration case data, wherein the target request type entity represents the arbitration request type of the labor arbitration case; and constructing a knowledge graph based on the target element entity, the target clause entity, the target reasoning template entity, and the target request type entity, wherein the knowledge graph is applied to the labor arbitration process.

Description

Technical Field

[0001] This disclosure relates to the field of knowledge graph technology, specifically to a method for constructing a knowledge graph, a method for applying a knowledge graph, and a corresponding device thereof. Background Technology

[0002] Labor arbitration aims to resolve labor disputes and places strict requirements and high expectations on the professional competence of arbitrators. Furthermore, the field of labor arbitration requires arbitrators to accumulate extensive experience through numerous real-world cases in order to continuously improve their professional skills, meticulously analyze labor disputes to uncover the truth, and render rulings that satisfy both parties.

[0003] However, there is currently a severe shortage of experienced arbitrators, and the gap between different regions is significant. For example, economically developed regions have a greater number of experienced arbitrators, while those in less developed regions tend to have less experience. Therefore, due to the varying levels of experience among arbitrators, the efficiency and reliability of labor arbitration are relatively low. Summary of the Invention

[0004] The purpose of this disclosure is to provide a method for constructing a knowledge graph, a method for applying the knowledge graph, and a corresponding device, in order to solve the problem that the efficiency and reliability of labor arbitration are low due to the different experience levels of arbitrators in related technologies.

[0005] To achieve the above objectives, a first aspect of this disclosure provides a method for constructing a knowledge graph, the method comprising:

[0006] Obtain data on labor arbitration cases;

[0007] Based on the preset target request type entity and the labor arbitration case data, the target element entity, target clause entity, and target reasoning template entity are determined. The target request type entity represents the type of arbitration request in the labor arbitration case.

[0008] A knowledge graph is constructed based on the target element entity, the target clause entity, the target reasoning template entity, and the target request type entity, and the knowledge graph is applied to the labor arbitration process.

[0009] Optionally, determining the target element entity based on the preset target request type entity and the labor arbitration case data includes:

[0010] Based on the target request type entity and the labor arbitration case data, determine the target evidence data corresponding to each target request type entity;

[0011] For each target request type entity, cluster analysis is performed on the target evidence data corresponding to the target request type entity to obtain multiple target evidence names, and the target evidence names that appear more frequently than a preset threshold among the multiple target evidence names are determined as the target element entities corresponding to the target request type entity.

[0012] Optionally, determining the target clause entity based on the preset target request type entity and the labor arbitration case data includes:

[0013] The target clause corresponding to each target request type entity is determined based on the target request type entity and the labor arbitration case data;

[0014] For each target request type entity, the target clause corresponding to the target request type entity is determined as the target clause entity corresponding to the target request type entity.

[0015] Optionally, determining the target reasoning template entity based on the preset target request type entity and the labor arbitration case data includes:

[0016] The labor arbitration case data is split according to the target request type entity to obtain the arbitration reasoning segment data corresponding to each target request type entity;

[0017] For each target request type entity, the arbitration reasoning segment data corresponding to the target request type entity is split into arbitration reasoning sentence data;

[0018] Cluster analysis is performed on the arbitration reasoning sentence data to obtain multiple target cluster centers, and the arbitration reasoning sentences represented by the multiple target cluster centers are determined as candidate sentence data;

[0019] Based on the candidate clause data, the target reasoning template entity corresponding to the target request type entity is determined.

[0020] Optionally, the cluster analysis of the arbitration reasoning clause data yields multiple target cluster centers, including:

[0021] The arbitration reasoning clause data is initialized to obtain multiple initial cluster centers;

[0022] Repeat the following clustering steps until the cluster centers of the arbitration reasoning clause data are fixed, and use the fixed cluster centers as the target cluster centers:

[0023] For each arbitration reasoning sentence in the arbitration reasoning sentence data, the similarity between the arbitration reasoning sentence and the cluster center is determined, and the arbitration reasoning sentence is assigned to the cluster corresponding to the cluster center with the highest similarity.

[0024] The similarity and sum of each arbitration reasoning clause in the cluster are determined with respect to other arbitration reasoning clauses in the cluster, and the arbitration reasoning clause with the highest similarity and sum of values ​​is taken as the new cluster center of the cluster.

[0025] Optionally, the initialization of the arbitration reasoning sentence data to obtain multiple initial cluster centers includes:

[0026] Determine a first probability value for each arbitration reasoning sentence in the arbitration reasoning sentence data, wherein the first probability value is positively correlated with the word frequency-reverse file frequency of the arbitration reasoning sentence;

[0027] The arbitration reasoning sentence with the highest first probability value in the arbitration reasoning sentence data is selected as the cluster center, and the following steps are repeated until the number of cluster centers equals the preset number:

[0028] Calculate the first distance between the arbitration reasoning clauses that are not cluster centers and the cluster centers in the arbitration reasoning clause data;

[0029] For each arbitration reasoning clause that did not become a cluster center, a second probability value is determined for the arbitration reasoning clause that did not become a cluster center. The second probability value is positively correlated with the minimum value of the first distances between the arbitration reasoning clauses that did not become cluster centers.

[0030] The arbitration reasoning sentence with the highest second probability value among the arbitration reasoning sentences that did not become cluster centers is taken as the cluster center.

[0031] Optionally, the step of constructing a knowledge graph based on the element entity, the target clause entity, the target reasoning template entity, and the target request type entity includes:

[0032] The first correspondence between the target request type entity and the target element entity is determined based on the request type entity and element entity included in the same labor arbitration case data;

[0033] A second correspondence between the target claim type entity and the target clause entity is determined based on the claim type entity and clause entity included in the same labor arbitration case data;

[0034] A third correspondence between the target request type entity and the target reasoning template entity is determined based on the request type entity and the reasoning template entity included in the same labor arbitration case data;

[0035] The fourth correspondence between the target element entity and the target clause entity is determined based on the element entity and clause entity included in the same labor arbitration case data;

[0036] The fifth correspondence between the target element entity and the reasoning template entity is determined based on the element entity and reasoning template entity included in the same labor arbitration case data;

[0037] The sixth correspondence between the target reasoning template entity and the target clause entity is determined based on the reasoning template entity and clause entity included in the same labor arbitration case data;

[0038] The knowledge graph is constructed based on the first correspondence, the second correspondence, the third correspondence, the fourth correspondence, the fifth correspondence, and the sixth correspondence.

[0039] A second aspect of this disclosure also provides a method for applying a knowledge graph, the method comprising:

[0040] Determine the type of arbitration claim in the case to be arbitrated;

[0041] Based on the arbitration request type and the preset knowledge graph, at least one of the following is determined: target candidate issue, target reference clause, and target reasoning template;

[0042] The knowledge graph is constructed in the following way:

[0043] Obtain data on labor arbitration cases;

[0044] Based on the preset target request type entity and the labor arbitration case data, the target element entity, target clause entity, and target reasoning template entity are determined. The target request type entity represents the type of arbitration request in the labor arbitration case.

[0045] A knowledge graph is constructed based on the target element entity, the target clause entity, the target reasoning template entity, and the target request type entity, and the knowledge graph is applied to the labor arbitration process.

[0046] Optionally, determining the target candidate problem based on the arbitration request type and a preset knowledge graph includes:

[0047] Based on the arbitration request type and the first correspondence between the target request type entity and the target element entity in the knowledge graph, the target element corresponding to the arbitration request type is determined, and the target candidate question is determined based on the target element.

[0048] Optionally, determining the target reference clause based on the arbitration request type and a preset knowledge graph includes:

[0049] Based on the arbitration request type and the second correspondence between the target request type entity and the target clause entity in the knowledge graph, the first clause corresponding to the arbitration request type is determined;

[0050] Based on the arbitration request type and the first correspondence between the target request type entity and the target element entity in the knowledge graph, the target element corresponding to the arbitration request type is determined, and based on the target element and the fourth correspondence between the target element entity and the target clause entity in the knowledge graph, the second clause corresponding to the target element is determined.

[0051] The target terms are determined in accordance with the first and second terms.

[0052] Optionally, determining the target reasoning template based on the arbitration request type and a preset knowledge graph includes:

[0053] Based on the arbitration request type and the third correspondence between the target request type entity and the target reasoning template entity in the knowledge graph, the first reasoning template corresponding to the arbitration request type is determined;

[0054] Based on the arbitration request type and the first correspondence between the target request type entity and the target element entity in the knowledge graph, the target element corresponding to the arbitration request type is determined, and based on the target element, the target element entity and the target reasoning template entity in the knowledge graph, the second reasoning template corresponding to the target element is determined.

[0055] The target reasoning template is determined based on the first reasoning template and the second reasoning template.

[0056] A third aspect of this disclosure also provides a knowledge graph construction apparatus, the apparatus comprising:

[0057] The acquisition module is used to acquire labor arbitration case data;

[0058] The first determining module is used to determine the target element entity, the target clause entity, and the target reasoning template entity based on the preset target request type entity and the labor arbitration case data. The target request type entity represents the type of arbitration request in the labor arbitration case.

[0059] A construction module is used to build a knowledge graph based on the target element entity, the target clause entity, the target reasoning template entity, and the target request type entity. The knowledge graph is applied to the labor arbitration process.

[0060] A fourth aspect of this disclosure also provides an application apparatus for a knowledge graph, the apparatus comprising:

[0061] The second determination module is used to determine the type of arbitration request in the case to be arbitrated;

[0062] The third determination module is used to determine at least one of the target candidate question, target reference clause, and target reasoning template based on the type of arbitration request and the preset knowledge graph.

[0063] The knowledge graph is constructed in the following way:

[0064] Obtain data on labor arbitration cases;

[0065] Based on the preset target request type entity and the labor arbitration case data, the target element entity, target clause entity, and target reasoning template entity are determined. The target request type entity represents the type of arbitration request in the labor arbitration case.

[0066] A knowledge graph is constructed based on the target element entity, the target clause entity, the target reasoning template entity, and the target request type entity, and the knowledge graph is applied to the labor arbitration process.

[0067] A fifth aspect of this disclosure also provides a non-transitory computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method described in any one of the first or second aspects above.

[0068] A sixth aspect of this disclosure also provides an electronic device, comprising:

[0069] A memory on which computer programs are stored;

[0070] A processor for executing the computer program in the memory to implement the steps of the method described in any one of the first or second aspects above.

[0071] The above technical solution can achieve at least the following technical effects:

[0072] This method involves acquiring labor arbitration case data and determining target element entities, target clause entities, and target reasoning template entities based on preset target request type entities and the data. A knowledge graph is then constructed based on these entities, and applied to the labor arbitration process. This method enables the construction of a labor arbitration knowledge graph, providing a standardized and unified descriptive framework for knowledge in the field. This knowledge graph can then be applied to the labor arbitration process, reducing the workload of manual intervention and avoiding limitations imposed by varying experience levels of arbitrators, thereby improving the efficiency and reliability of labor arbitration.

[0073] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0074] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings:

[0075] Figure 1 This is a flowchart of a method for constructing a knowledge graph according to an exemplary embodiment;

[0076] Figure 2 This is a flowchart of an arbitration reasoning clause clustering analysis provided according to an exemplary embodiment;

[0077] Figure 3 This is a schematic diagram of an arbitration reasoning template provided according to an exemplary embodiment;

[0078] Figure 4 This is a schematic diagram of a subgraph of a knowledge graph provided according to an exemplary embodiment;

[0079] Figure 5 This is a schematic diagram illustrating a knowledge graph construction process according to an exemplary embodiment;

[0080] Figure 6 This is a flowchart of a knowledge graph application method provided according to an exemplary embodiment;

[0081] Figure 7 This is a block diagram of a knowledge graph construction apparatus provided according to an exemplary embodiment;

[0082] Figure 8 This is a block diagram of a knowledge graph application device provided according to an exemplary embodiment;

[0083] Figure 9 This is a block diagram of an electronic device provided according to an exemplary embodiment. Detailed Implementation

[0084] The specific embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit this disclosure.

[0085] It should be understood that the various steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect. The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Relevant definitions for other terms will be given in the description below.

[0086] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules, or units, and are not used to limit the order of functions performed by these devices, modules, or units or their interdependencies. It should also be noted that the modifications of "a" and "a plurality of" mentioned in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0087] Labor arbitration is an important component of the human resources and social security system and the preferred method for resolving labor disputes. A sound labor arbitration system can quickly and effectively protect vulnerable groups in critical moments, effectively safeguard the legitimate rights and interests of workers and employers, and reduce unnecessary civil disputes arising from labor disputes.

[0088] However, there is currently a severe shortage of experienced arbitrators, and the gap between different regions is significant. For example, economically developed regions have a greater number of experienced arbitrators, while those in less developed regions tend to have less experience. Therefore, the efficiency and reliability of labor arbitration are limited by the varying levels of experience among arbitrators. Furthermore, how to efficiently and quickly transfer the rich experience of experienced arbitrators to less experienced ones has become a crucial issue that urgently needs to be addressed in the field of arbitration.

[0089] This disclosure provides a method for constructing and applying a knowledge graph, as well as a corresponding apparatus, to solve the above-mentioned problems.

[0090] The following provides a detailed description of the embodiments of the technical solution disclosed herein.

[0091] Reference Figure 1 This disclosure provides a method for constructing a knowledge graph, the method comprising:

[0092] S101. Obtain labor arbitration case data.

[0093] S102. Determine the target element entity, target clause entity, and target reasoning template entity based on the preset target request type entity and labor arbitration case data.

[0094] Among them, the target request type entity represents the type of arbitration request in a labor arbitration case.

[0095] S103. Construct a knowledge graph based on the target element entity, target clause entity, target reasoning template entity, and target request type entity.

[0096] Knowledge graphs have been applied to the labor arbitration process.

[0097] Using the above method, labor arbitration case data is acquired. Based on the preset target request type entities and the labor arbitration case data, target element entities, target clause entities, and target reasoning template entities are determined. Then, a knowledge graph is constructed based on these entities, which is applied to the labor arbitration process. This method allows for the construction of a labor arbitration knowledge graph, providing a standardized and unified descriptive framework for knowledge in the field of labor arbitration. This knowledge graph can then be applied to the labor arbitration process, reducing the workload of manual intervention and avoiding limitations imposed by varying experience levels of arbitrators, thereby improving the efficiency and reliability of labor arbitration.

[0098] To enable those skilled in the art to better understand the knowledge graph construction method provided in this disclosure, the above steps are illustrated in detail below.

[0099] Before constructing the knowledge graph, labor arbitration cases are analyzed and summarized to design the structure of the knowledge graph. This disclosed embodiment, based on the concept of element-based case handling, proposes a knowledge graph in the field of labor arbitration. It constructs element entities based on the types of claims in arbitration cases, then mines relevant policies and regulations from big data texts, and finally infers and summarizes reasoning templates.

[0100] First, a labor arbitration case typically includes one or more arbitration requests, which are clearly stated in the application when the case is filed, such as "Request the respondent to pay the applicant's medical expenses of XXXX yuan" or "Request confirmation that an employment relationship exists between the applicant and the respondent." The arbitration institution needs to process each request reasonably and legally. In this process, the request type is an important reference point; similar types of arbitration requests tend to share commonalities, such as the applicability of the same policies and regulations. Therefore, request type entities can be used as the basis for a knowledge graph.

[0101] Secondly, given the complexity of labor arbitration cases, even cases of the same type that share certain commonalities can still have many sub-categories. For example, even with arbitration requests of the "determination of employment relationship" type, whether the applicant has evidence such as bank statements showing wage payments from the respondent could lead to completely different arbitration results. Therefore, it is possible to define essential entities based on the type of request—key information that allows for more granular differentiation of case circumstances. For example, "pay slips," "bank statements," and "attendance records" are all essential entities related to the "determination of employment relationship" entity.

[0102] Then, since labor arbitration awards need to be based on law and verifiable evidence, the award must directly cite laws and regulations to form effective reasoning paragraphs, thereby leading to the award result. Therefore, this publicly available knowledge graph defines clause entities to correspond to specific policy and regulatory clauses.

[0103] Finally, in the drafting of labor arbitration awards, it is necessary to explain the reasons for the award by stating the facts and legal basis. Therefore, the reasoning section is a very important component of the award. Furthermore, this reasoning section requires arbitrators to have extensive experience in skillfully applying the relevant laws and regulations. Therefore, this published knowledge graph defines a reasoning template entity. Through big data analysis, reasoning fragments can be extracted from labor arbitration case data and combined into reasoning templates. This yields the main body of the reasoning section for cases with different claim types and specific elements, with the remaining small portion needing to be filled in based on the specific case. In actual cases, arbitrators only need to fill in the blanks based on the appropriate reasoning template. Since the legal provisions and main reasoning statements are provided by the template, the goal of using a knowledge graph to fill in domain experience is achieved.

[0104] Furthermore, after identifying the request type entity, element entity, clause entity, and reasoning template entity, the relationships between the request type entity and the element entity, the request type entity and the clause entity, the request type entity and the reasoning template entity, the element entity and the clause entity, the element entity and the reasoning template entity, and the clause entity and the reasoning template entity are defined, thereby constructing a knowledge graph.

[0105] The relationships between request type entities and element entities are as follows: The request type entity points to the element entity, indicating the key elements that may be included in the request type. The request type entity points to the clause entity, indicating the policy and regulatory clauses that may be referenced in the request type. The request type entity points to the reasoning template entity, indicating the template that may be used in the reasoning section of the ruling for this request type. The element entity points to the clause entity, indicating the policy and regulatory clauses that may be referenced when that element appears. The element entity points to the reasoning template entity, indicating the template that may be used in the reasoning section of the ruling when that element appears. The clause entity points to the reasoning template entity, indicating the policy and regulatory clauses referenced in that reasoning template.

[0106] The following describes how to construct the knowledge graph of an embodiment of this disclosure.

[0107] It should be noted that the labor arbitration case data includes documents generated from previous labor arbitrations, such as applications, awards, and hearing transcripts; this disclosure does not impose any restrictions on this. Furthermore, data from labor arbitration cases handled by senior arbitrators can be selected as data for constructing the knowledge graph to improve its quality.

[0108] For example, firstly, a set of request type entities is established. From the perspective of labor arbitration, the scope of request types is relatively fixed and the boundaries are relatively clear. Relevant departments typically have general guidance documents on labor arbitration request types, dividing them into eight major categories. Then, different provinces or regions further subdivide the labor arbitration request types based on these eight categories. In other words, the target request type entity set can be clearly formed directly based on these guidance documents. Specific adjustments can be made according to the actual guidance documents; this disclosure does not impose any restrictions on this.

[0109] For example, the above "requesting the respondent to pay the applicant's medical expenses of XXXX yuan" is an arbitration request under the subcategory of "work-related injury medical expenses" within the subcategory of "work-related injury medical treatment and compensation" under the major category of "social insurance disputes," while "requesting confirmation of the existence of an employment relationship between the applicant and the respondent" is a request under the category of "dispute over the determination of employment relationship."

[0110] In one possible approach, determining the target element entity based on the preset target request type entity and labor arbitration case data can be as follows: determine the target evidence data corresponding to each target request type entity based on the target request type entity and labor arbitration case data; for each target request type entity, perform cluster analysis on the target evidence data corresponding to the target request type entity to obtain multiple target evidence names; and determine the target evidence names whose frequency of occurrence exceeds a preset threshold among the multiple target evidence names as the target element entity corresponding to the target request type entity.

[0111] For example, elements, as key information that distinguishes different case scenarios of the same request type at a finer granular level, play a crucial role in the selection of reasoning templates and even other higher-level application scenarios. In actual case trials, the evidence provided by both parties greatly influences the determination of facts and the outcome of the judgment. Therefore, in this embodiment of the disclosure, element entities can be extracted from the case evidence list.

[0112] For example, if the evidence list cannot be directly provided from structured data, it can be extracted from the judgment or court transcript. Generally, judgments and court transcripts have a certain organizational structure, clearly indicating the evidence provided by both the applicant and the respondent. Below is an excerpt from a judgment in a case: "To prove the facts asserted, the applicant submits the following evidence to this Commission: Evidence 1: Employment contract, proving that an employment relationship exists between the applicant and the respondent and the applicant's salary amount; Evidence 2: Payroll records, proving the applicant's salary amount." Therefore, evidence such as "employment contract" and "payroll records" can be extracted from the judgment through template matching or other methods to form the evidence list for the case.

[0113] For example, by analyzing and mining a large number of arbitration awards or court transcripts, evidence lists are extracted for each case, forming an arbitration case evidence set, based on different claim types. Then, using metrics such as edit distance and longest common subsequence as distance measures, cluster analysis is used to unify similar evidence expressions. The most frequent expression or the expression closest to the category center is selected as the unique expression for that element category; that is, one element category corresponds to one expression form (evidence name). Finally, element categories with a frequency exceeding a certain preset threshold are selected to form an element entity set. It should be noted that the element entities corresponding to different claim types may be different.

[0114] In one possible approach, determining the target clause entity based on the preset target request type entity and labor arbitration case data could be as follows: determining the target clause corresponding to each target request type entity based on the target request type entity and labor arbitration case data, and for each target request type entity, determining the target clause corresponding to the target request type entity as the target clause entity corresponding to the target request type entity.

[0115] For example, policy and regulatory provisions serve as the theoretical basis for the ruling and directly reflect the persuasiveness of the judgment. The reasoning and conclusion sections of the ruling explicitly cite relevant policy and regulatory provisions, and the patterns of these provisions are usually quite fixed, such as: "According to Articles 7 and 44 of the XX Law, ...". Therefore, by using template matching and other methods, the cited policy and regulatory provisions can be extracted from the reasoning and conclusion sections of the ruling for different types of claims. By analyzing and mining a large number of ruling texts, a set of clause entities can be obtained, and then the target clause entities corresponding to different claim types can be identified.

[0116] In one possible approach, determining the target reasoning template entity based on the preset target request type entity and labor arbitration case data can be as follows: The labor arbitration case data is split according to the target request type entity to obtain arbitration reasoning segment data corresponding to each target request type entity. For each target request type entity, the arbitration reasoning segment data corresponding to the target request type entity is split into arbitration reasoning sentence data. Cluster analysis is performed on the arbitration reasoning sentence data to obtain multiple target cluster centers. The arbitration reasoning sentences represented by the multiple target cluster centers are determined as candidate sentence data. The target reasoning template entity corresponding to the target request type entity is determined based on the candidate sentence data.

[0117] For example, reasoning templates provide arbitrators with a basic framework for their arguments, requiring less experience from them while providing strong evidence for their decisions. High-quality awards have a structured approach, from basic case information, the applicant's claims and the respondent's responses, to the evidence presented and examined by both parties, the arbitral tribunal's opinions on the admissibility of the evidence, and finally, the arbitral conclusion reached through reasoning. Each part is logically clear and interconnected, making the overall award well-reasoned, persuasive, and well-supported.

[0118] Therefore, in this embodiment of the disclosure, the arbitration reasoning section is extracted by dissecting the structure of the arbitration award. Typically, a well-organized arbitration reasoning section will provide a segmented explanation for each arbitration claim made by the applicant, specifying which claim each segment corresponds to. For example: "...(arbitration reasoning section), therefore, the applicant's first arbitration claim for compensation for unlawful termination of the employment relationship is without legal basis and is not supported by this commission." In this way, the arbitration claims and their corresponding reasoning paragraphs can be obtained. By analyzing and mining a large amount of arbitration award text, key materials and a preliminary form of a set of reasoning template entities can be constructed.

[0119] For example, the process can be divided into the following steps: First, aggregate the reasoning texts of the same request type to form a reasoning set for each request type (arbitration reasoning segment data). Then, for each request type, split the reasoning within its set into sentences using punctuation marks such as commas (arbitration reasoning sentence data). Next, perform cluster analysis on the sentence units to retain high-frequency sentences that exceed a certain threshold (candidate sentence data). Finally, find several reasoning paragraphs containing these high-frequency sentences from the reasoning set, and together with the high-frequency sentences, form the reasoning template material and prototype for this type of labor arbitration request.

[0120] One possible approach to clustering the arbitration reasoning sentence data to obtain multiple target cluster centers is as follows: Initialize the arbitration reasoning sentence data to obtain multiple initial cluster centers, and repeat the following clustering steps until the cluster centers of the arbitration reasoning sentence data are fixed, and use the fixed cluster centers as target cluster centers: For each arbitration reasoning sentence in the arbitration reasoning sentence data, determine the similarity between the arbitration reasoning sentence and the cluster center, and assign the arbitration reasoning sentence to the cluster corresponding to the cluster center with the highest similarity. Determine the similarity and sum of each arbitration reasoning sentence in the cluster and other arbitration reasoning sentences in the cluster, and use the arbitration reasoning sentence with the highest similarity and sum of similarity as the new cluster center of the cluster.

[0121] For example, the above clustering steps can be performed using the overall process of the traditional K-Mediods clustering algorithm. However, arbitration reasoning clauses have certain special characteristics, and directly applying the traditional K-Mediods clustering algorithm will not yield good results. Therefore, this disclosure improves the traditional K-Mediods clustering algorithm based on actual conditions.

[0122] For example, the K-Mediods clustering algorithm is initialized randomly, that is, K sentences are randomly selected as the initial cluster centers with uniform probability. The quality of the initialization directly affects the effect of K-Mediods clustering, and the clustering results desired by the embodiments of this disclosure are to satisfy the following two tendencies: (1) the semantic differences between different clusters should be large, that is, under a fixed K value, we should try to obtain reasoning materials with different semantics; (2) we should be able to clearly cluster important reasoning materials containing actual content into each category, rather than focusing too much on certain fixed-form clichés. Therefore, in the initialization process, the initial K cluster centers should satisfy the above two tendencies as much as possible.

[0123] In one possible approach, initializing the arbitration reasoning sentence data to obtain multiple initial cluster centers can be achieved by: determining a first probability value corresponding to each arbitration reasoning sentence in the arbitration reasoning sentence data, where the first probability value is positively correlated with the term frequency-reverse document frequency of the arbitration reasoning sentence; selecting the arbitration reasoning sentence with the highest first probability value in the arbitration reasoning sentence data as the cluster center; and repeating the following steps until the number of cluster centers equals a preset number: calculating the first distance between the arbitration reasoning sentences that did not become cluster centers and the cluster centers; for each arbitration reasoning sentence that did not become a cluster center, determining a second probability value corresponding to the arbitration reasoning sentence that did not become a cluster center, where the second probability value is positively correlated with the minimum of the first distances of the arbitration reasoning sentences that did not become cluster centers; and selecting the arbitration reasoning sentence with the highest second probability value in the arbitration reasoning sentence data that did not become a cluster center as the cluster center.

[0124] For example, the initialization process of the K-Means++ algorithm is introduced in the initialization section to address the design where there are significant semantic differences between different clusters. Simultaneously, a new distance calculation formula for arbitration reasoning sentences is proposed, combining the TF-IDF (Term Frequency-Inverse Document Frequency) index of words, enabling the clear clustering of important reasoning materials containing actual content into each category. Here, TF-IDF aims to evaluate the importance of words to the corpus, which is a collection of reasoning texts corresponding to different request types.

[0125] It's worth noting that K-Mediods clustering algorithms typically use Euclidean distance or cosine similarity for distance calculation. However, traditional Euclidean distance or cosine similarity calculations treat words indiscriminately, failing to differentiate the varying importance of different words. In practical applications, such as the aforementioned element entities and clause entities, when two argumentative clauses co-occur with entities from a knowledge graph, such as citing the same policy or regulatory clause, they are more similar than those co-occurring with ordinary words. Furthermore, when entities co-occur, the text length of the entity also affects similarity; longer entities are more explicit in meaning, and the argumentative clauses are more semantically similar. Therefore, based on traditional cosine similarity, we can introduce entity sets from a knowledge graph and comprehensively consider entity text length and TF-IDF indices to optimize the similarity calculation formula. This allows the clustering algorithm to better suit the specific characteristics of arbitration argumentative clauses, thereby achieving better results.

[0126] The clustering analysis process of embodiments of this disclosure is illustrated in detail below.

[0127] Reference Figure 2 This disclosure provides a clustering analysis process for arbitration reasoning clauses, the process including:

[0128] S201. For the arbitration reasoning clause data, select a clause s with probability p1 as the first cluster center.

[0129] Wherein, probability p1 is proportional to the value of clause s. Value, of which Indicates w i The tf-idf metric, w i The words in s.

[0130] S202. Calculate the distance between clause c, which has not yet become a cluster center, and clause s, which has become a cluster center.

[0131] For example, the distance calculation formula is as follows:

[0132]

[0133] Where d represents distance, cossim() represents traditional cosine similarity, and c i Let be the words in clause c, and α be a hyperparameter with a value of [0,1].

[0134] S203. Select a clause c that has not yet become a cluster center with probability p2 as the N+1th cluster center.

[0135] Where the probability p2 is proportional to min{d(c, s)} i )} (minimum value), where s i ∈S, where S is the current set of cluster centers, and |S|=N.

[0136] Further, repeat steps S202 and S203 above until there is a set of cluster centers |S|=K, that is, all K cluster centers have been determined. At this point, the initialization process is complete. The determination of the value of K can be referred to the process of determining the value of K in the K-Means++ algorithm in related technologies, and will not be repeated here.

[0137] S204. For the arbitration reasoning clause data, calculate the similarity between each clause x and each cluster center clause s, and assign each clause x to the cluster represented by the center with the highest similarity.

[0138] For example, the similarity calculation formula is as follows:

[0139]

[0140] Where len() represents the word length, y i For knowledge graph entities that appear simultaneously in cluster center clauses s and x.

[0141] S205. For each cluster H, assume each clause h in the cluster as the centroid of the cluster, calculate the similarity of other clauses in the cluster to the assumed centroid, and sum them. , where h i ∈Hh, and finally select the clause h with the maximum sum of similarity as the new center point of the cluster.

[0142] S206. Determine whether the center point of each cluster H has changed.

[0143] Furthermore, if the center point (clustering result) of each cluster H changes, repeat steps S204 and S205. If the clustering result no longer changes, the clustering process is considered to be officially completed, and step S207 is executed.

[0144] S207. Take the center point of each cluster H as the target cluster center, and determine the arbitration reasoning clauses represented by multiple target cluster centers as candidate clause data.

[0145] For example, through automatic mining and analysis of a large amount of arbitration award texts, reasoning template materials and prototypes (candidate clause data) can be obtained. These can then be manually reviewed and assembled by business experts to ensure the correctness and seriousness of all reasoning, thus forming a formal set of arbitration reasoning template entities, such as... Figure 3 The arbitration reasoning template shown is a template that meets the key elements of "workers' compensation insurance" under the "workers' medical expenses" claim type, with the underlined parts being reserved blanks. Of course, in other possible ways, artificial intelligence algorithms can also be used for automatic assembly, and this disclosure does not limit this.

[0146] In one possible approach, constructing a knowledge graph based on element entities, target clause entities, target reasoning template entities, and target request type entities can be as follows: First, determine the first correspondence between target request type entities and target element entities based on the request type entities and element entities included in the same labor arbitration case data; second, determine the second correspondence between target request type entities and target clause entities based on the request type and clause entities included in the same labor arbitration case data; third, determine the third correspondence between target request type entities and target reasoning template entities based on the request type entities and reasoning template entities included in the same labor arbitration case data; fourth, determine the fourth correspondence between request target element entities and target clause entities based on the element entities and clause entities included in the same labor arbitration case data; fifth, determine the sixth correspondence between target element entities and reasoning template entities based on the element entities and reasoning template entities included in the same labor arbitration case data; and construct a knowledge graph based on the first, second, third, fourth, fifth, and sixth correspondences.

[0147] For example, in the process of establishing entities from the above-mentioned data mining and analysis of labor arbitration cases, the relationships between the entities are also basically established. For instance, when establishing element entities, the element entities mined from the same labor arbitration case data (such as arbitration awards) can be related to the claim types in that labor arbitration case data. When establishing clause entities, the clause entities mined from the same labor arbitration case data can be related to the claim types in that labor arbitration case data. When establishing arbitration reasoning template entities, they have already been established separately according to each claim type, so the arbitration reasoning template entity can be directly related to its corresponding claim type. When establishing element entities and clause entities, the element entities and clause entities mined from the same labor arbitration case data are related. In establishing the entity of the arbitration reasoning template, it is necessary to extract arbitration reasoning sentences from the award and form high-frequency reasoning sentence clusters through cluster analysis. Then, for each arbitration reasoning sentence used in the final arbitration reasoning template, the element entities appearing in the source award of each sentence within its cluster are extracted. By analyzing a large number of award texts, high-frequency element entities that appear more than a certain threshold are retained and their relationships are established with that arbitration reasoning template. Furthermore, based on the relevant policy and regulatory clauses directly cited in the arbitration reasoning template, the relationship between the arbitration reasoning template entity and the clause entity is established. A knowledge graph is then built based on these relationships, referencing... Figure 4 Here is an example of a subgraph of a knowledge graph.

[0148] Thus, referring to Figure 5This system can establish relationships between entities based on guidance documents and labor arbitration case data, resulting in a knowledge graph for the labor arbitration field. It can fill gaps in labor arbitration knowledge graphs and accurately describe and quickly convey the rich domain knowledge and experience of senior arbitrators, assisting newly appointed arbitrators in handling labor disputes. Furthermore, it proposes a standardized and unified descriptive framework for labor arbitration knowledge, thereby eliminating the problem of different standards among arbitrators. It also fully leverages the advantages of information technology, greatly reducing the workload of traditional manual intervention while summarizing domain knowledge more comprehensively and richly.

[0149] Based on the same inventive concept, this disclosure provides a method for applying a knowledge graph, referring to... Figure 6 The method includes:

[0150] S601. Determine the type of arbitration request in the case to be arbitrated.

[0151] S602. Determine at least one of the following based on the type of arbitration request and the pre-set knowledge graph: target candidate issue, target reference clause, and target reasoning template.

[0152] The knowledge graph is constructed in the following way: acquiring labor arbitration case data; determining target element entities, target clause entities, and target reasoning template entities based on preset target request type entities and labor arbitration case data, with the target request type entity representing the type of arbitration request in the labor arbitration case; constructing a knowledge graph based on the target element entities, target clause entities, target reasoning template entities, and target request type entities, and applying the knowledge graph to the labor arbitration process.

[0153] By applying this method to the labor arbitration process, knowledge graphs can be standardized and unified, breaking the experience limitations of arbitrators, reducing the workload of manual intervention, and improving the efficiency and reliability of labor arbitration.

[0154] In one possible approach, the target candidate question is determined based on the type of arbitration request and a pre-defined knowledge graph, including: determining the target element corresponding to the type of arbitration request based on the type of arbitration request and the first correspondence between the request type entity and the element entity in the knowledge graph, and determining the target candidate question based on the target element.

[0155] For example, before labor arbitration, a basic list of candidate hearing questions can be generated based on the type of arbitration request in the case. This list can be used by arbitrators for selection and reference, preparing for the hearing. For instance, based on the type of arbitration request (corresponding to the request type entity in the knowledge graph), element entities related to that type of arbitration request can be found in the knowledge graph. Then, using these elements as the core, some fixed-pattern questions can be assembled to obtain candidate hearing questions. Using the "attendance record" element entity as the core, hearing questions for the applicant and respondent can be obtained such as: "Does the applicant have attendance records? If so, from what year, month, and day to what year, month, and day?" "What are the respondent's opinions on the attendance records and time periods provided by the applicant?", etc. This disclosure does not impose any limitations on this.

[0156] In one possible approach, the target reference clause is determined based on the type of arbitration request and a pre-defined knowledge graph, including: determining the first clause corresponding to the type of arbitration request based on the second correspondence between the type of arbitration request and the entity of the request type in the knowledge graph; determining the target element corresponding to the type of arbitration request based on the first correspondence between the type of arbitration request and the entity of the request type in the knowledge graph; determining the second clause corresponding to the target element based on the fourth correspondence between the target element and the entity of the element in the knowledge graph; and determining the target clause based on the first clause and the second clause.

[0157] For example, policy and regulatory provisions serve as the legal basis for the arbitration award and are crucial from the initial assessment of the case to the drafting of the award. Therefore, based on the type of arbitration claim and the key elements gradually identified as the trial progresses, a knowledge graph can be used to determine the first clause corresponding to the type of arbitration claim and the second clause corresponding to the target element. These first and second clauses can then be presented to the arbitrator as the target clauses for reference in the legal basis of the case. Alternatively, clauses with a high degree of relevance to both the claim type entity and the element entity can be found in the knowledge graph and presented to the arbitrator as a reference in the legal basis of the case; that is, the target clauses are determined from the first and second clauses. This disclosure does not impose any restrictions on this approach.

[0158] In one possible approach, the target reasoning template is determined based on the type of arbitration request and a pre-defined knowledge graph. This includes: determining a first reasoning template corresponding to the type of arbitration request based on the third correspondence between the type of arbitration request, the entity of the request type in the knowledge graph, and the entity of the reasoning template; determining a target element corresponding to the type of arbitration request based on the first correspondence between the type of arbitration request, the entity of the request type in the knowledge graph, and the entity of the element in the knowledge graph, and the entity of the reasoning template based on the fifth correspondence between the target element, the entity of the element in the knowledge graph, and the entity of the reasoning template; and finally determining the target reasoning template based on the first and second reasoning templates.

[0159] For example, as the final formal written document presented to the applicant and respondent in a labor arbitration case, the arbitration award must have a logically clear and persuasive reasoning section to gain acceptance of the award. Therefore, when drafting the award, based on the type of arbitration claim, the key elements involved in the case can be determined by comprehensively considering the evidence presented by both parties and the court proceedings. Using a knowledge graph, a first reasoning template corresponding to the type of arbitration claim and a second reasoning template corresponding to the target elements can be identified. These first and second reasoning templates can then be used as target reasoning templates and provided to the arbitrators as candidate frameworks for the reasoning paragraphs of the award for reference and supplementation. Alternatively, arbitration reasoning template entities with high relevance to both the claim type entity and the element entity can be found in the knowledge graph and used as candidate frameworks for the reasoning paragraphs of the award for reference and supplementation by the arbitrators. That is, the target reasoning template can be determined from the first and second reasoning templates; this disclosure does not impose any limitations on this approach.

[0160] It should be noted that the above-mentioned application scenarios of knowledge graphs are merely illustrative examples. In practical applications, knowledge graphs can be involved in the entire process of case handling, from pre-trial preparation to the drafting of the judgment. This disclosure does not impose any restrictions on this.

[0161] Based on the same inventive concept, embodiments of this disclosure provide a knowledge graph construction apparatus, referring to... Figure 7 The construction apparatus includes:

[0162] Module 701 is used to acquire data on labor arbitration cases.

[0163] The first determining module 702 is used to determine the target element entity, the target clause entity, and the target reasoning template entity based on the preset target request type entity and the labor arbitration case data. The target request type entity represents the arbitration request type of the labor arbitration case.

[0164] The construction module 703 is used to construct a knowledge graph based on the target element entity, the target clause entity, the target reasoning template entity, and the target request type entity. The knowledge graph is applied to the labor arbitration process.

[0165] Using this device, a knowledge graph for labor arbitration can be constructed, which provides a standardized and unified descriptive framework for knowledge in the field of labor arbitration. This knowledge graph can then be applied to the labor arbitration process, reducing the workload of manual intervention and avoiding the limitations imposed by the different experience levels of arbitrators in related technologies, thereby improving the efficiency and reliability of labor arbitration.

[0166] Optionally, the first determining module 702 is used to:

[0167] Based on the target request type entity and the labor arbitration case data, determine the target evidence data corresponding to each target request type entity;

[0168] For each target request type entity, cluster analysis is performed on the target evidence data corresponding to the target request type entity to obtain multiple target evidence names, and the target evidence names that appear more frequently than a preset threshold among the multiple target evidence names are determined as the target element entities corresponding to the target request type entity.

[0169] Optionally, the first determining module 702 is used to:

[0170] The target clause corresponding to each target request type entity is determined based on the target request type entity and the labor arbitration case data;

[0171] For each target request type entity, the target clause corresponding to the target request type entity is determined as the target clause entity corresponding to the target request type entity.

[0172] Optionally, the first determining module 702 is used to:

[0173] The labor arbitration case data is split according to the target request type entity to obtain the arbitration reasoning segment data corresponding to each target request type entity;

[0174] For each target request type entity, the arbitration reasoning segment data corresponding to the target request type entity is split into arbitration reasoning sentence data;

[0175] Cluster analysis is performed on the arbitration reasoning sentence data to obtain multiple target cluster centers, and the arbitration reasoning sentences represented by the multiple target cluster centers are determined as candidate sentence data;

[0176] Based on the candidate clause data, the target reasoning template entity corresponding to the target request type entity is determined.

[0177] Optionally, the first determining module 702 is used to:

[0178] The arbitration reasoning clause data is initialized to obtain multiple initial cluster centers;

[0179] Repeat the following clustering steps until the cluster centers of the arbitration reasoning clause data are fixed, and use the fixed cluster centers as the target cluster centers:

[0180] For each arbitration reasoning sentence in the arbitration reasoning sentence data, the similarity between the arbitration reasoning sentence and the cluster center is determined, and the arbitration reasoning sentence is assigned to the cluster corresponding to the cluster center with the highest similarity.

[0181] The similarity and sum of each arbitration reasoning clause in the cluster are determined with respect to other arbitration reasoning clauses in the cluster, and the arbitration reasoning clause with the highest similarity and sum of values ​​is taken as the new cluster center of the cluster.

[0182] Optionally, the first determining module 702 is used to:

[0183] Determine a first probability value for each arbitration reasoning sentence in the arbitration reasoning sentence data, wherein the first probability value is positively correlated with the word frequency-reverse file frequency of the arbitration reasoning sentence;

[0184] The arbitration reasoning sentence with the highest first probability value in the arbitration reasoning sentence data is selected as the cluster center, and the following steps are repeated until the number of cluster centers equals the preset number:

[0185] Calculate the first distance between the arbitration reasoning clauses that are not cluster centers and the cluster centers in the arbitration reasoning clause data;

[0186] For each arbitration reasoning clause that did not become a cluster center, a second probability value is determined for the arbitration reasoning clause that did not become a cluster center. The second probability value is positively correlated with the minimum value of the first distances between the arbitration reasoning clauses that did not become cluster centers.

[0187] The arbitration reasoning sentence with the highest second probability value among the arbitration reasoning sentences that did not become cluster centers is taken as the cluster center.

[0188] Optionally, the building module 703 is used for:

[0189] The first correspondence between the target request type entity and the target element entity is determined based on the request type entity and element entity included in the same labor arbitration case data;

[0190] A second correspondence between the target claim type entity and the target clause entity is determined based on the claim type entity and clause entity included in the same labor arbitration case data;

[0191] A third correspondence between the target request type entity and the target reasoning template entity is determined based on the request type entity and the reasoning template entity included in the same labor arbitration case data;

[0192] The fourth correspondence between the target element entity and the target clause entity is determined based on the element entity and clause entity included in the same labor arbitration case data;

[0193] The fifth correspondence between the target element entity and the reasoning template entity is determined based on the element entity and reasoning template entity included in the same labor arbitration case data;

[0194] The sixth correspondence between the target reasoning template entity and the target clause entity is determined based on the reasoning template entity and clause entity included in the same labor arbitration case data;

[0195] The knowledge graph is constructed based on the first correspondence, the second correspondence, the third correspondence, the fourth correspondence, the fifth correspondence, and the sixth correspondence.

[0196] Based on the same inventive concept, this disclosure also provides an application device for knowledge graphs, referring to... Figure 8 The application device includes:

[0197] The second determining module 801 is used to determine the type of arbitration request in the case to be arbitrated.

[0198] The third determining module 802 is used to determine at least one of the target candidate issue, target reference clause, and target reasoning template based on the arbitration request type and the preset knowledge graph.

[0199] The knowledge graph is constructed in the following way:

[0200] Obtain data on labor arbitration cases;

[0201] Based on the preset target request type entity and the labor arbitration case data, the target element entity, target clause entity, and target reasoning template entity are determined. The target request type entity represents the type of arbitration request in the labor arbitration case.

[0202] A knowledge graph is constructed based on the target element entity, the target clause entity, the target reasoning template entity, and the target request type entity, and the knowledge graph is applied to the labor arbitration process.

[0203] By using this device to apply knowledge graphs to the labor arbitration process, the process can be standardized and unified, breaking the experience limitations of arbitrators, reducing the workload of manual intervention, and improving the efficiency and reliability of labor arbitration.

[0204] Optionally, the third determining module 802 is used to:

[0205] Based on the arbitration request type and the first correspondence between the target request type entity and the target element entity in the knowledge graph, the target element corresponding to the arbitration request type is determined, and the target candidate question is determined based on the target element.

[0206] Optionally, the third determining module 802 is used to:

[0207] Based on the arbitration request type and the second correspondence between the target request type entity and the target clause entity in the knowledge graph, the first clause corresponding to the arbitration request type is determined;

[0208] Based on the arbitration request type and the first correspondence between the target request type entity and the target element entity in the knowledge graph, the target element corresponding to the arbitration request type is determined, and based on the target element and the fourth correspondence between the target element entity and the target clause entity in the knowledge graph, the second clause corresponding to the target element is determined.

[0209] The target terms are determined in accordance with the first and second terms.

[0210] Optionally, the third determining module 802 is used to:

[0211] Based on the arbitration request type and the third correspondence between the target request type entity and the target reasoning template entity in the knowledge graph, the first reasoning template corresponding to the arbitration request type is determined;

[0212] Based on the arbitration request type and the first correspondence between the target request type entity and the target element entity in the knowledge graph, the target element corresponding to the arbitration request type is determined, and based on the target element, the target element entity and the target reasoning template entity in the knowledge graph, the second reasoning template corresponding to the target element is determined.

[0213] The target reasoning template is determined based on the first reasoning template and the second reasoning template.

[0214] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0215] Based on the same inventive concept, this disclosure provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of any of the methods described in the above-mentioned knowledge graph construction or application methods.

[0216] Based on the same inventive concept, this disclosure provides an electronic device, including:

[0217] A memory on which computer programs are stored;

[0218] A processor for executing the computer program in the memory to implement the steps of any of the methods described above for constructing or applying the knowledge graph.

[0219] Figure 9 This is a block diagram illustrating an electronic device 900 according to an exemplary embodiment, with reference to... Figure 9 The electronic device 900 includes a processor 901, which may be one or more, and a memory 902 for storing computer programs executable by the processor 901. The computer programs stored in the memory 902 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processor 901 may be configured to execute the computer program to perform the aforementioned knowledge graph construction or application methods.

[0220] Additionally, the electronic device 900 may also include a power supply component 905 and a communication component 903. The power supply component 905 can be configured to perform power management of the electronic device 900, and the communication component 903 can be configured to enable communication of the electronic device 900, such as wired or wireless communication. Furthermore, the electronic device 900 may also include an input / output (I / O) interface 904. The electronic device 900 can operate on an operating system, such as Windows Server, stored in the memory 902. TM Mac OSX TM Unix TM Linux TM etc.

[0221] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the knowledge graph construction method or application method described above. For example, the non-transitory computer-readable storage medium may be the memory 902 including program instructions described above, which may be executed by the processor 901 of the electronic device 900 to complete the knowledge graph construction method or application method described above.

[0222] In another exemplary embodiment, a computer program product is also provided, the computer program product comprising a computer program executable by a programmable device, the computer program having a code portion for performing the above-described knowledge graph construction method or application method when executed by the programmable device.

[0223] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.

[0224] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0225] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

Claims

1. A method for constructing a knowledge graph, characterized in that, The method includes: Obtain data on labor arbitration cases; Based on the preset target request type entity and the labor arbitration case data, the target element entity, target clause entity, and target reasoning template entity are determined. The target request type entity represents the type of arbitration request in the labor arbitration case. A knowledge graph is constructed based on the target element entity, the target clause entity, the target reasoning template entity, and the target request type entity, and the knowledge graph is applied to the labor arbitration process. The target element entity is an entity related to the target evidence data corresponding to the target request type entity; In constructing the knowledge graph, the relationships between target request type entities and target element entities, target request type entities and target clause entities, target request type entities and target reasoning template entities, target element entities and target clause entities, target element entities and target reasoning template entities, and target clause entities and target reasoning template entities are defined to construct the knowledge graph. Specifically, the relationship between a target request type entity and a target element entity points from the target request type entity to the target element entity, indicating that the target request type contains key elements; the relationship between a target request type entity and a target clause entity points from the target request type entity to the target clause entity, indicating that the target request type references key elements. The policy and regulatory provisions; the relationship between the target request type entity and the target reasoning template entity is that the target request type entity points to the target reasoning template entity, indicating that the template to be applied in the reasoning section of the ruling in this target request type; the relationship between the target element entity and the target clause entity is that the target element entity points to the target clause entity, indicating that the policy and regulatory provisions cited when the target element appears; the relationship between the target element entity and the target reasoning template entity is that the target element entity points to the target reasoning template entity, indicating that the template to be applied in the reasoning section of the ruling when the target element appears; the relationship between the target clause entity and the target reasoning template entity is that the target reasoning template entity points to the target clause entity, indicating that the policy and regulatory provisions cited in this target reasoning template.

2. The method according to claim 1, characterized in that, The step of determining the target element entity based on the preset target request type entity and the labor arbitration case data includes: Based on the target request type entity and the labor arbitration case data, determine the target evidence data corresponding to each target request type entity; For each target request type entity, cluster analysis is performed on the target evidence data corresponding to the target request type entity to obtain multiple target evidence names, and the target evidence names that appear more frequently than a preset threshold among the multiple target evidence names are determined as the target element entities corresponding to the target request type entity.

3. The method according to claim 1, characterized in that, The step of determining the target clause entity based on the preset target request type entity and the labor arbitration case data includes: The target clause corresponding to each target request type entity is determined based on the target request type entity and the labor arbitration case data; For each target request type entity, the target clause corresponding to the target request type entity is determined as the target clause entity corresponding to the target request type entity.

4. The method according to claim 1, characterized in that, The step of determining the target reasoning template entity based on the preset target request type entity and the labor arbitration case data includes: The labor arbitration case data is split according to the target request type entity to obtain the arbitration reasoning segment data corresponding to each target request type entity; For each target request type entity, the arbitration reasoning segment data corresponding to the target request type entity is split into arbitration reasoning sentence data; Cluster analysis is performed on the arbitration reasoning sentence data to obtain multiple target cluster centers, and the arbitration reasoning sentences represented by the multiple target cluster centers are determined as candidate sentence data; Based on the candidate clause data, the target reasoning template entity corresponding to the target request type entity is determined.

5. The method according to claim 4, characterized in that, The cluster analysis of the arbitration reasoning clause data yields multiple target cluster centers, including: The arbitration reasoning clause data is initialized to obtain multiple initial cluster centers; Repeat the following clustering steps until the cluster centers of the arbitration reasoning clause data are fixed, and use the fixed cluster centers as the target cluster centers: For each arbitration reasoning sentence in the arbitration reasoning sentence data, the similarity between the arbitration reasoning sentence and the cluster center is determined, and the arbitration reasoning sentence is assigned to the cluster corresponding to the cluster center with the highest similarity. The similarity and sum of each arbitration reasoning clause in the cluster are determined with respect to other arbitration reasoning clauses in the cluster, and the arbitration reasoning clause with the highest similarity and sum of values ​​is taken as the new cluster center of the cluster.

6. The method according to claim 5, characterized in that, The initialization of the arbitration reasoning sentence data yields multiple initial cluster centers, including: Determine a first probability value for each arbitration reasoning sentence in the arbitration reasoning sentence data, wherein the first probability value is positively correlated with the word frequency-reverse file frequency of the arbitration reasoning sentence; The arbitration reasoning sentence with the highest first probability value in the arbitration reasoning sentence data is selected as the cluster center, and the following steps are repeated until the number of cluster centers equals the preset number: Calculate the first distance between the arbitration reasoning clauses that are not cluster centers and the cluster centers in the arbitration reasoning clause data; For each arbitration reasoning clause that did not become a cluster center, a second probability value is determined for the arbitration reasoning clause that did not become a cluster center. The second probability value is positively correlated with the minimum value of the first distances between the arbitration reasoning clauses that did not become cluster centers. The arbitration reasoning sentence with the highest second probability value among the arbitration reasoning sentences that did not become cluster centers is taken as the cluster center.

7. The method according to any one of claims 1-6, characterized in that, The construction of a knowledge graph based on the element entity, the target clause entity, the target reasoning template entity, and the target request type entity includes: The first correspondence between the target request type entity and the target element entity is determined based on the request type entity and element entity included in the same labor arbitration case data; A second correspondence between the target claim type entity and the target clause entity is determined based on the claim type entity and clause entity included in the same labor arbitration case data; A third correspondence between the target request type entity and the target reasoning template entity is determined based on the request type entity and the reasoning template entity included in the same labor arbitration case data; The fourth correspondence between the target element entity and the target clause entity is determined based on the element entity and clause entity included in the same labor arbitration case data; The fifth correspondence between the target element entity and the reasoning template entity is determined based on the element entities and reasoning template entities included in the same labor arbitration case data; The sixth correspondence between the target reasoning template entity and the target clause entity is determined based on the reasoning template entity and clause entity included in the same labor arbitration case data; The knowledge graph is constructed based on the first correspondence, the second correspondence, the third correspondence, the fourth correspondence, the fifth correspondence, and the sixth correspondence.

8. A method for applying knowledge graphs, characterized in that, The method includes: Determine the type of arbitration claim in the case to be arbitrated; Based on the arbitration request type and the preset knowledge graph, at least one of the following is determined: target candidate issue, target reference clause, and target reasoning template; The knowledge graph is constructed in the following way: Obtain data on labor arbitration cases; Based on the preset target request type entity and the labor arbitration case data, the target element entity, target clause entity, and target reasoning template entity are determined. The target request type entity represents the type of arbitration request in the labor arbitration case. A knowledge graph is constructed based on the target element entity, the target clause entity, the target reasoning template entity, and the target request type entity, and the knowledge graph is applied to the labor arbitration process. The target element entity is an entity related to the target evidence data corresponding to the target request type entity; In constructing the knowledge graph, the relationships between target request type entities and target element entities, target request type entities and target clause entities, target request type entities and target reasoning template entities, target element entities and target clause entities, target element entities and target reasoning template entities, and target clause entities and target reasoning template entities are defined to construct the knowledge graph. Specifically, the relationship between a target request type entity and a target element entity points from the target request type entity to the target element entity, indicating that the target request type contains key elements; the relationship between a target request type entity and a target clause entity points from the target request type entity to the target clause entity, indicating that the target request type references key elements. The policy and regulatory provisions; the relationship between the target request type entity and the target reasoning template entity is that the target request type entity points to the target reasoning template entity, indicating that the template to be applied in the reasoning section of the ruling in this target request type; the relationship between the target element entity and the target clause entity is that the target element entity points to the target clause entity, indicating that the policy and regulatory provisions cited when the target element appears; the relationship between the target element entity and the target reasoning template entity is that the target element entity points to the target reasoning template entity, indicating that the template to be applied in the reasoning section of the ruling when the target element appears; the relationship between the target clause entity and the target reasoning template entity is that the target reasoning template entity points to the target clause entity, indicating that the policy and regulatory provisions cited in this target reasoning template.

9. The method according to claim 8, characterized in that, The step of determining the target candidate problem based on the arbitration request type and a preset knowledge graph includes: Based on the arbitration request type and the first correspondence between the target request type entity and the target element entity in the knowledge graph, the target element corresponding to the arbitration request type is determined, and the target candidate question is determined based on the target element.

10. The method according to claim 8, characterized in that, The step of determining the target reference clause based on the arbitration request type and a pre-defined knowledge graph includes: Based on the arbitration request type and the second correspondence between the target request type entity and the target clause entity in the knowledge graph, the first clause corresponding to the arbitration request type is determined; Based on the arbitration request type and the first correspondence between the target request type entity and the target element entity in the knowledge graph, the target element corresponding to the arbitration request type is determined, and based on the target element and the fourth correspondence between the target element entity and the target clause entity in the knowledge graph, the second clause corresponding to the target element is determined. The target terms are determined in accordance with the first and second terms.

11. The method according to claim 8, characterized in that, The step of determining the target reasoning template based on the arbitration request type and a preset knowledge graph includes: Based on the arbitration request type and the third correspondence between the target request type entity and the target reasoning template entity in the knowledge graph, the first reasoning template corresponding to the arbitration request type is determined; Based on the arbitration request type and the first correspondence between the target request type entity and the target element entity in the knowledge graph, the target element corresponding to the arbitration request type is determined, and based on the target element, the target element entity and the target reasoning template entity in the knowledge graph, the second reasoning template corresponding to the target element is determined. The target reasoning template is determined based on the first reasoning template and the second reasoning template.

12. A knowledge graph construction apparatus, characterized in that, The device includes: The acquisition module is used to acquire labor arbitration case data; The first determining module is used to determine the target element entity, the target clause entity, and the target reasoning template entity based on the preset target request type entity and the labor arbitration case data. The target request type entity represents the type of arbitration request in the labor arbitration case. The construction module is used to build a knowledge graph based on the target element entity, the target clause entity, the target reasoning template entity, and the target request type entity. The knowledge graph is applied to the labor arbitration process. The target element entity is an entity related to the target evidence data corresponding to the target request type entity; In constructing the knowledge graph, the relationships between target request type entities and target element entities, target request type entities and target clause entities, target request type entities and target reasoning template entities, target element entities and target clause entities, target element entities and target reasoning template entities, and target clause entities and target reasoning template entities are defined to construct the knowledge graph. Specifically, the relationship between a target request type entity and a target element entity points from the target request type entity to the target element entity, indicating that the target request type contains key elements; the relationship between a target request type entity and a target clause entity points from the target request type entity to the target clause entity, indicating that the target request type references key elements. The policy and regulatory provisions; the relationship between the target request type entity and the target reasoning template entity is that the target request type entity points to the target reasoning template entity, indicating that the template to be applied in the reasoning section of the ruling in this target request type; the relationship between the target element entity and the target clause entity is that the target element entity points to the target clause entity, indicating that the policy and regulatory provisions cited when the target element appears; the relationship between the target element entity and the target reasoning template entity is that the target element entity points to the target reasoning template entity, indicating that the template to be applied in the reasoning section of the ruling when the target element appears; the relationship between the target clause entity and the target reasoning template entity is that the target reasoning template entity points to the target clause entity, indicating that the policy and regulatory provisions cited in this target reasoning template.

13. A knowledge graph application device, characterized in that, The device includes: The second determination module is used to determine the type of arbitration request in the case to be arbitrated; The third determination module is used to determine at least one of the target candidate question, target reference clause, and target reasoning template based on the type of arbitration request and the preset knowledge graph. The knowledge graph is constructed in the following way: Obtain data on labor arbitration cases; Based on the preset target request type entity and the labor arbitration case data, the target element entity, target clause entity, and target reasoning template entity are determined. The target request type entity represents the type of arbitration request in the labor arbitration case. A knowledge graph is constructed based on the target element entity, the target clause entity, the target reasoning template entity, and the target request type entity, and the knowledge graph is applied to the labor arbitration process. The target element entity is an entity related to the target evidence data corresponding to the target request type entity; In constructing the knowledge graph, the relationships between target request type entities and target element entities, target request type entities and target clause entities, target request type entities and target reasoning template entities, target element entities and target clause entities, target element entities and target reasoning template entities, and target clause entities and target reasoning template entities are defined to construct the knowledge graph. Specifically, the relationship between a target request type entity and a target element entity points from the target request type entity to the target element entity, indicating that the target request type contains key elements; the relationship between a target request type entity and a target clause entity points from the target request type entity to the target clause entity, indicating that the target request type references key elements. The policy and regulatory provisions; the relationship between the target request type entity and the target reasoning template entity is that the target request type entity points to the target reasoning template entity, indicating that the template to be applied in the reasoning section of the ruling in this target request type; the relationship between the target element entity and the target clause entity is that the target element entity points to the target clause entity, indicating that the policy and regulatory provisions cited when the target element appears; the relationship between the target element entity and the target reasoning template entity is that the target element entity points to the target reasoning template entity, indicating that the template to be applied in the reasoning section of the ruling when the target element appears; the relationship between the target clause entity and the target reasoning template entity is that the target reasoning template entity points to the target clause entity, indicating that the policy and regulatory provisions cited in this target reasoning template.

14. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method described in any one of claims 1-11.

15. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-11.

Images