WO2020048058A1 - 基金知识推理方法、系统、计算机设备和存储介质 - Google Patents

基金知识推理方法、系统、计算机设备和存储介质 Download PDF

Info

Publication number
WO2020048058A1
WO2020048058A1 PCT/CN2018/124220 CN2018124220W WO2020048058A1 WO 2020048058 A1 WO2020048058 A1 WO 2020048058A1 CN 2018124220 W CN2018124220 W CN 2018124220W WO 2020048058 A1 WO2020048058 A1 WO 2020048058A1
Authority
WO
WIPO (PCT)
Prior art keywords
fund knowledge
fund
node
knowledge
metabase
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2018/124220
Other languages
English (en)
French (fr)
Inventor
陈泽晖
胡逸凡
黄鸿顺
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ping An Technology Shenzhen Co Ltd
Original Assignee
Ping An Technology Shenzhen Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Ping An Technology Shenzhen Co Ltd filed Critical Ping An Technology Shenzhen Co Ltd
Publication of WO2020048058A1 publication Critical patent/WO2020048058A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/22Indexing; Data structures therefor; Storage structures
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24Querying
    • G06F16/245Query processing
    • G06F16/2455Query execution
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/04Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02DCLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00Energy efficient computing, e.g. low power processors, power management or thermal management

Definitions

  • the present application relates to the field of graph theory and network analysis technology, and in particular, to methods, systems, computer equipment, and storage media for fund knowledge reasoning.
  • centrality is an indicator that determines the importance of nodes in a network, and it is a quantification of the importance of nodes.
  • These centrality indicators are initially applied in social networks.
  • people and their connections are modeled into a network graph. Nodes in the network graph represent people, and connecting edges between nodes represent people-to-people connections. Based on the established network structure diagram, a series of centrality measurement methods can be used to calculate which individual is more important than other individuals.
  • the current knowledge graph is only a relational network display composed of knowledge points connected to each other.
  • the relational network is composed of nodes and node relationships.
  • the importance of the existing technology to a node to other nodes cannot be performed well. Multi-angle and dynamic analysis.
  • a fund knowledge reasoning method includes: extracting fund knowledge from an information source in a fund knowledge reasoning platform to establish a fund knowledge metabase, where the fund knowledge metabase includes multiple subgraphs, each of which includes entities, relationships, and Attributes; fuse the fund knowledge in the fund knowledge metabase according to preset rules; store the fused fund knowledge metabase in a database; enter a query request at the front desk of the fund knowledge inference platform, and The query request is sent to the background of the fund knowledge inference platform; the sub-graph corresponding to the query request is selected from the fund knowledge metabase, and the node centrality of the entity in the sub-graph is calculated, and the node The calculation result of the centrality is returned to the front desk for display; the subgraph corresponding to the query request is received, and the shortest path between the two entities in the subgraph is calculated, and the shortest path between the two entities is calculated according to Calculate the node centrality of each entity.
  • a fund knowledge inference system includes: an extraction unit configured to extract fund knowledge from an information source in a fund knowledge inference platform to establish a fund knowledge metabase, and the fund knowledge metabase includes multiple sub-graphs, each sub-graph Including entities, relationships, and attributes; a fusion unit configured to fuse the fund knowledge in the fund knowledge metabase according to a preset rule; a storage unit configured to store the fused fund knowledge metabase in a database
  • the input unit is configured to input a query request at the front desk of the fund knowledge inference platform and send the query request to the backstage of the fund knowledge inference platform; the query unit is configured to filter out the relevant information from the fund knowledge metabase.
  • the subgraph corresponding to the query request is calculated, and the node centrality of the entity in the subgraph is calculated, and the calculation result of the node centrality is returned to the front desk for display; the operation unit is configured to receive the query request associated with the query request.
  • Corresponding subgraph calculate the shortest path between two entities in the subgraph, and according to the shortest path between the two entities Calculation of the center nodes of each entity.
  • a computer device includes a memory and a processor.
  • the memory stores computer-readable instructions.
  • the processor causes the processor to execute the steps of the fund knowledge inference method described above.
  • a storage medium storing computer-readable instructions.
  • the processor causes one or more of the processors to execute the steps of the fund knowledge inference method described above.
  • the above-mentioned fund knowledge inference method, system, computer equipment and storage medium by extracting fund knowledge and establishing a fund knowledge metabase; merging fund knowledge in the fund knowledge metabase and storing it in a database; into the fund knowledge metabase Filter out the subgraph corresponding to the query request, calculate the node centrality of each entity in the subgraph, and return the calculation result of the node centrality to the foreground for display; calculate the shortest path between two entities in the subgraph, and according to The shortest path between two entities is used to calculate the node centrality of each entity.
  • the real-time computing platform uses an analysis framework that includes a centrality measurement algorithm to perform computational analysis on the query request, and realizes multi-angle and dynamic analysis of each node in the full graph or each sub-graph. Centrality analysis improves work efficiency.
  • FIG. 1 is a flowchart of a fund knowledge reasoning method in an embodiment of the present application
  • FIG. 3 is a flowchart of a graph query process in an embodiment of the present application.
  • FIG. 5 is a structural block diagram of a fund knowledge inference system in an embodiment of the present application.
  • FIG. 1 is a flowchart of a fund knowledge reasoning method provided in an embodiment of the application, as shown in the figure, including: S1, extracting fund knowledge from an information source in a fund knowledge reasoning platform to establish a fund knowledge metabase,
  • the fund knowledge metabase contains multiple subgraphs, each of which includes entities, relationships, and attributes.
  • the knowledge contained in the information source is extracted through processes such as identification, understanding, screening, and induction to establish a knowledge metabase.
  • the knowledge metabase includes entities, relationships, and attributes.
  • the entity performs ID identification.
  • the fusion process includes the fusion of new data to replace the old data, and also includes evaluation of the quality of knowledge and fusion with weights according to preset fusion rules.
  • S3. The fused knowledge base of the fund is stored in a database. In this step, the fused data is stored.
  • the storage database may be a relational database, an RDF database, a graph database, or any combination of databases. the way.
  • S4. Enter a query request at the front desk of the fund knowledge inference platform, and send the query request to the back desk of the fund knowledge inference platform. In this step, a query request is input at the front desk of the fund knowledge inference platform.
  • the query request includes the user's need for funds
  • the data of A the front desk transmits the query request containing fund A to the backstage of the fund knowledge inference platform.
  • the sub-graph corresponding to the query request is selected from the fund knowledge metabase, and the node centrality of the entity in the sub-graph is calculated, and the calculation result of the node centrality is returned to the front desk.
  • the backstage in this step filters the sub-graph corresponding to fund A into the fund knowledge metabase according to the keywords of fund A included in the query request, and transmits the corresponding sub-graph to the fund knowledge
  • the node centrality of each entity is calculated in the real-time operator of the inference platform, the background obtains the calculation result from the real-time operator, and returns the calculation result to the front desk in json data format.
  • the front desk uses d3js technology displays the calculation results. S6.
  • the real-time operator described above obtains the subgraph, assigns the edge weights according to the entity relationship type according to a preset rule, and then calculates the shortest path between the two entities through a preset algorithm, and according to the calculated shortest path between the two entities.
  • This application implements the calculation function of the centrality of each node in the knowledge graph through the above steps and methods. Through this calculation, it is possible to analyze the importance of a node to other nodes in the graph from multiple angles and dynamics, which is more helpful for mining the knowledge graph. Potential relationships between nodes in the.
  • FIG. 2 is a flowchart of a graph construction process provided in an embodiment of the present application, as shown in the figure, including: S101, identifying fund knowledge contained in an information source, and identifying data types and data of the fund knowledge Source; in this step, the knowledge in the fund knowledge metabase is identified according to its data type and data source.
  • the data in the internal database of the enterprise is structured data
  • the chart data in websites such as Tiantian Fund Network is semi-structured data.
  • the entire text of fund research reports, fund manager resumes, and snowball community reviews are unstructured data.
  • the merging the fund knowledge in the fund knowledge metabase according to a preset rule includes: ID identification of each entity in the fund knowledge metabase; in this step, Before the knowledge data in the knowledge metabase is fused according to a preset fusion rule, ID identification is performed on all entities, for example, a fund entity and a stock entity use a market transaction code as the ID identification. Judging each entity in the fund knowledge metabase, the same entity with a unified ID identifier is the same entity, and the same entity is merged with the relationship and attributes, if it does not have a unified ID identifier, according to the similarity of the attributes of each entity Merge.
  • the fusion of data in this step includes the fusion of new data with the replacement of old data, and also includes the evaluation of the quality of knowledge and the fusion with weights according to a preset fusion rule, which is an entity in the fund knowledge metabase. Perform ID identification, fuse the relationship and attributes for entities identified by the same ID, and fuse similar attributes for entities that do not have the same ID.
  • ID identification is performed on the entities in the fund knowledge metabase, and then the ID identification entities are fused according to a preset fusion rule, and the knowledge in the knowledge metabase is integrated in an orderly manner. Subsequently, the required fund information of the fund can be quickly found in the fund knowledge metabase.
  • the method further includes regularly checking and updating entities, relationships, and attributes in the knowledge base of funds, which are regularly checked and updated. Including checking whether the entity and the relationship in the fund knowledge metabase have changed within a fixed period of time, and if the entity or the relationship changes, the changed entity and the relationship are changed Update to the Fund Knowledge Metabase.
  • the database in this step includes a relational database, an RDF database, a graph database, or a combination of any of them.
  • the periodic inspection and update includes checking the entities and institutions in the fund knowledge metabase within a fixed period of time. Whether there is a change in the relationship, and if the entity or the relationship changes, the changed entity and the relationship are updated to the fund knowledge metabase.
  • FIG. 3 is a flowchart of a graph query process provided in an embodiment of the present application.
  • the flowchart includes: S501. Receive the query request, and send the query request to the query according to keywords of the query request. The corresponding sub-graph is matched in the fund knowledge metabase.
  • the background receives the query request sent by the front desk. For example, if the user needs to query the fund manager with the widest / most core contacts in the scope, the back office obtains the query.
  • the requested keywords for example, the query condition is who is the most well-connected fund manager among Fudan University alumni, who is the most core fund manager under E Fund Fund Company, and matches the corresponding subgraph in the fund knowledge metabase . S502.
  • the calculation results of the node centrality of each sub-earth are obtained, for example, the calculation results of the sub-graph of Fudan University and the sub-graph of E Fund Fund are obtained, and the calculation result of the sub-graph that best meets the query request is returned to the front desk. S505.
  • the result of the node centrality is returned to the front desk in a json data format, and the result of the node centrality is displayed on the front desk by using d3js technology.
  • the process of matching the query request to the corresponding subgraph helps to quickly locate the graph relationship closest to the query request, and helps to quickly perform node centrality on the entities included in the query request. analysis.
  • FIG. 4 is a flowchart of a real-time calculation process provided in an embodiment of the present application.
  • the flowchart includes calculation of a weighted multi-source shortest path, and calculation of the weighted multi-source shortest path includes The weights of edges are assigned according to the entity relationship type according to a preset rule, and then a multi-source shortest path is calculated by a preset algorithm; S601.
  • the multi-source shortest path includes the shortest paths of any two or two entities in the sub-graph; in this step, According to the preset rules according to the preset rules, for example, the relationship between fund experiences is weighted.
  • the weight is set Is 1; if the relationship between the two fund managers is graduated from the same school, the same company or classmates, the weight is set to 2. S602. At the same time as calculating the shortest path between two entities in the sub-graph, the calculation also includes the degree of degree centrality, which includes the relationship of one of the entities in the sub-graph. quantity.
  • the shortest path between two entities in the sub-graph is calculated according to a preset rule, which provides a basis for subsequent calculation of the node centrality of each entity in the sub-graph.
  • a subgraph corresponding to the query request is received, a shortest path between two entities in the subgraph is calculated, and a node center of each entity is calculated according to the shortest path between the two entities.
  • the degree further includes: obtaining a calculation result of the multi-source shortest path of any two or two entities in the sub-graph; and substituting the calculation result of the multi-source shortest path into the formula (1) and the formula (2) respectively to calculate the intermediary of the corresponding node Centrality and compact centrality, where formula (1) and formula (2) are as follows:
  • Formula (1) represents the intermediary centrality of node i in the subgraph, where Pjk represents the number of shortest paths between any two nodes jk, and Pjk (i) is the number of shortest paths between node jk and node i.
  • the intermediary centrality includes the number of times a node serves as the shortest path between the other two nodes, the higher the number, the greater the intermediary node degree;
  • formula (2) represents the closeness of node x in the subgraph Centrality, d (y, x) represents the length of the shortest path from node x to any node y, that is, the tight centrality of node x is the reciprocal of the sum of the shortest path distances from x to all other nodes, and the tight centrality includes The degree to which a node is located at the center of the network in the subgraph. If a node is close to many other nodes, the closer the node is to the center of the network; the intermediary centrality and the close center of the node The degree is summed, and the result of the sum is the node centrality.
  • the median degree of centrality and tightness of the entities in the sub-graph are calculated. Sum of degrees and tight centrality finally get the calculation result of the node centrality of the corresponding entity, which realizes the multi-angle and dynamic analysis of the centrality of each node in the full graph and each sub-graph, which improves the work efficiency.
  • the fund knowledge inference system includes an extraction unit, a fusion unit, a storage unit, an input unit, a query unit, and an operation unit, of which: An extraction unit is set up to extract fund knowledge from an information source in a fund knowledge inference platform to establish a fund knowledge metabase, where the fund knowledge metabase includes multiple subgraphs, each of which includes entities, relationships, and attributes; a fusion unit , Configured to fuse the fund knowledge in the fund knowledge metabase according to a preset rule; a storage unit configured to store the fused fund knowledge metabase in a database; an input unit configured to be stored in a fund The front desk of the knowledge inference platform inputs a query request, and sends the query request to the background of the fund knowledge inference platform; the query unit is configured to filter the sub-graph corresponding to the query request to the fund knowledge metabase, and Calculate the node centrality of the entity in the sub-graph, and return the calculation result
  • the fusion unit includes: an identification module configured to identify each entity in the fund knowledge metabase; a merging module configured to determine each entity in the fund knowledge metabase , The entity with the unified ID identifier is the same entity, and the same entity is merged with the relationship and the attributes, and the entity without the unified ID identifier is merged according to the similarity of the attributes of the entities.
  • the fund knowledge metabase further includes a checking module configured to periodically check and update entities, relationships, and attributes in the fund knowledge metabase, and to check the fund knowledge metabase within a fixed period of time. Whether there is a change in the entity and the relationship in, and if the entity or the relationship changes, the changed entity and the relationship are updated to the fund knowledge metabase.
  • the query unit includes: a receiving request module, configured to receive the query request, and match the corresponding subgraph to the fund knowledge metabase according to the keywords of the query request; and obtain the centrality A module configured to calculate a node centrality of one or more of the subgraphs included in the keyword; a return module configured to obtain a result of the node centrality of each of the subgraphs, and will be most consistent with the The result of querying the node centrality is returned to the front desk; the result of the node centrality is returned to the front desk in json data format, and the result of the node centrality is displayed on the front desk using d3js technology. .
  • the operation unit includes a calculation module configured to calculate a shortest path with weights and multiple sources, and the calculation of the shortest path with weights includes allocating edge weights according to entity relationship types according to preset rules, and then The multi-source shortest path is calculated by a preset algorithm; the multi-source shortest path includes the shortest paths of any two or two entities in the subgraph; while the shortest paths between the two or two entities in the subgraph are calculated, Including the calculation of the degree centrality; the degree centrality includes the number of the relations of a certain entity in the sub-graph.
  • the calculation module further includes: obtaining a path result, configured to obtain a calculation result of a multi-source shortest path of any two or two entities in the sub-graph; a first-level calculation module, configured to set the multi-source The calculation result of the shortest path is substituted into the formula to calculate the intermediary centrality and tight centrality of the corresponding node respectively; the secondary calculation module is set to perform a sum operation on the intermediary centrality and the tight centrality of the node, and the calculation The result is the node centrality.
  • a computer device in one embodiment, includes a memory and a processor.
  • the memory stores computer-readable instructions.
  • the processor causes the processor to perform the foregoing operations. Steps of the fund knowledge reasoning method in the embodiment.
  • a storage medium storing computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, one or more of the processors execute the foregoing embodiments. Steps in the Fundamental Knowledge Reasoning Method.
  • the storage medium may be a non-volatile storage medium.
  • the program may be stored in a computer-readable storage medium.
  • the storage medium may include: Read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), magnetic disks or optical disks, etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Data Mining & Analysis (AREA)
  • Economics (AREA)
  • Operations Research (AREA)
  • Game Theory and Decision Science (AREA)
  • General Business, Economics & Management (AREA)
  • Quality & Reliability (AREA)
  • Development Economics (AREA)
  • Software Systems (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Computational Linguistics (AREA)
  • Stored Programmes (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)

Abstract

一种基金知识推理方法、系统、计算机设备和存储介质,涉及软件开发和维护技术领域。基金知识推理方法包括:抽取基金知识并建立基金知识元库;对基金知识元库中的基金知识进行融合并将其存储于数据库中;到基金知识元库中筛选出与查询请求对应的子图,计算子图中各实体的节点中心度,将节点中心度的计算结果返回至前台进行展示;计算子图中两两实体间的最短路径,并根据所述两两实体间的最短路径来计算每个实体的节点中心度。该方法通过基金知识图谱的构建与维护,实现了以多角度、动态地对图谱中各节点的中心性分析功能,提高了工作效率。

Description

基金知识推理方法、系统、计算机设备和存储介质
本申请要求于2018年09月03日提交中国专利局、申请号为201811019068.9、发明名称为“基金知识推理方法、系统、计算机设备和存储介质”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本申请涉及图论与网络分析技术领域,特别是涉及基金知识推理方法、系统、计算机设备和存储介质。
背景技术
在图论与网络分析中,中心性是判定网络中节点重要性的指标,是节点重要性的量化,这些中心性度量指标最初应用在社会网络中。在社会网络中,将人以及人与人之间的联系被模型化成网络图,网络图中的节点代表人,节点之间的连边表示人与人之间的联系。基于建立起来的网络结构图,使用一系列中心性度量方法就可以计算出哪个个体比其他个体更重要。
目前的知识图谱,只是由知识点相互连接而成的关系网络展示,该关系网络都是由节点和节点关系构成,但是现有技术对某个节点对其他节点的重要程度,不能很好地进行多角度和动态分析。
发明内容
基于此,有必要针对现有技术不能很好地对关系网络中某个节点对其它节点的重要程度进行多角度的动态分析的问题,提供一种基金知识推理方法、系统、计算机设备和存储介质。
一种基金知识推理方法,包括:抽取基金知识推理平台中的信息源中的基金知识后建立基金知识元库,所述基金知识元库中包含多个 子图,各子图中包括实体、关系和属性;根据预设规则将所述基金知识元库中的所述基金知识进行融合;将经过融合的所述基金知识元库存储于数据库中;在基金知识推理平台的前台输入查询请求,并将所述查询请求发送至基金知识推理平台的后台;到所述基金知识元库中筛选出与所述查询请求对应的子图,并计算该子图中的实体的节点中心度,将所述节点中心度的计算结果返回至所述前台进行展示;接收与所述查询请求对应的子图,计算所述子图中两两实体间的最短路径,并根据所述两两实体间的最短路径来计算每个实体的节点中心度。
一种基金知识推理系统,包括:抽取单元,设置为抽取基金知识推理平台中的信息源中的基金知识后建立基金知识元库,所述基金知识元库中包含多个子图,各子图中包括实体、关系和属性;融合单元,设置为根据预设规则将所述基金知识元库中的所述基金知识进行融合;存储单元,设置为将经过融合的所述基金知识元库存储于数据库中;输入单元,设置为在基金知识推理平台的前台输入查询请求,并将所述查询请求发送至基金知识推理平台的后台;查询单元,设置为到所述基金知识元库中筛选出与所述查询请求对应的子图,并计算该子图中的实体的节点中心度,将所述节点中心度的计算结果返回至所述前台进行展示;运算单元,设置为接收与所述查询请求的对应的子图,计算所述子图中两两实体间的最短路径,并根据所述两两实体间的最短路径来计算每个实体的节点中心度。
一种计算机设备,包括存储器和处理器,所述存储器中存储有计算机可读指令,所述计算机可读指令被所述处理器执行时,使得所述处理器执行上述基金知识推理方法的步骤。
一种存储有计算机可读指令的存储介质,所述计算机可读指令被一个或多个处理器执行时,使得一个或多个所述处理器执行如上述基金知识推理方法的步骤。
上述基金知识推理方法、系统、计算机设备和存储介质,通过抽取基金知识并建立基金知识元库;对基金知识元库中的基金知识进行融合并将其存储于数据库中;到基金知识元库中筛选出与查询请求对应的子图,计算子图中各实体的节点中心度,将节点中心度的计算结果返回至前台进行展示;计算子图中两两实体间的最短路径,并根据所述两两实体间的最短路径来计算每个实体的节点中心度。因此,对于输入的某一查询请求,实时运算平台运用一种包含中心性度量算法的分析框架对该查询请求进行计算分析,实现了多角度、动态地对全图谱或各子图中各节点的中心性分析,提高了工作效率。
附图说明
通过阅读下文优选实施方式的详细描述,各种其他的优点和益处对于本领域普通技术人员将变得清楚明了。附图仅用于示出优选实施方式的目的,而并不认为是对本申请的限制。
图1为本申请在一个实施例中基金知识推理方法的流程图;
图2为本申请在一个实施例中图谱构建过程的流程图;
图3为本申请在一个实施例中图谱查询过程的流程图;
图4为本申请在一个实施例中实施运算过程的流程图;
图5为本申请在一个实施例中基金知识推理系统的结构框图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本申请进行进一步详细说明。应当理解,此处所描 述的具体实施例仅仅用以解释本申请,并不用于限定本申请。
图1为本申请在一个实施例中提供的基金知识推理方法的流程图,如图所示,包括:S1、抽取基金知识推理平台中的信息源中的基金知识后建立基金知识元库,所述基金知识元库中包含多个子图,各子图中包括实体、关系和属性;本步骤通过把蕴含于信息源中的知识经过识别、理解、筛选、归纳等过程抽取出来,建立知识元库,所述知识元库包括实体、关系和属性。S2、根据预设规则将所述基金知识元库中的所述基金知识进行融合;本步骤通过使来自不同知识源的知识在同一框架规范下进行数据整合,并对所述知识元库中的实体进行ID标识,该融合过程中包括新数据替换旧数据的融合,还包括根据预设融合规则对知识的质量进行评估和带权重的融合。S3、将经过融合的所述基金知识元库存储于数据库中;本步骤将经过融合处理后的数据进行存储,存储数据库可以采用关系数据库,RDF数据库,图数据库等,或者采用任意数据库相结合的方式。S4、在基金知识推理平台的前台输入查询请求,并将所述查询请求发送至基金知识推理平台的后台;本步骤通过在基金知识推理平台的前台输入查询请求,所述查询请求包括用户需要基金A的数据,所述前台将包含了基金A的所述查询请求传送至基金知识推理平台的后台。S5、到所述基金知识元库中筛选出与所述查询请求对应的子图,并计算该子图中的实体的节点中心度,将所述节点中心度的计算结果返回至所述前台进行展示;本步骤中所述后台根据所述查询请求中包含的基金A的关键词到所述基金知识元库中筛选出与基金A对应的子图,并将该对应的子图传送中基金知识推理平台的实时运算器中进行各实体的节点中心度的计算,所述后台从所述实时运算器中获取该计算结果,将该计 算结果以json数据格式返回至所述前台,所述前台运用d3js技术将该计算结果进行展示。S6、接收与所述查询请求对应的子图,计算所述子图中两两实体间的最短路径,并根据所述两两实体间的最短路径来计算每个实体的节点中心度;本步骤中所述实时运算器获取所述子图,根据预设规则按实体关系类型分配边的权重,再通过预设算法计算两两实体间的最短路径,根据计算得到的两两实体间的最短路径,再根据预设算法计算出所述子图中各节点的中介中心度和紧密中心的,将子图中各节点的中介中心度和紧密中心度进行求和运算,得到该节点的节点中心度。
本申请通过上述步骤方法,实现了对知识图谱中各节点中心度的计算功能,通过该计算能够多角度和动态的分析图谱中某个节点对其它节点的重要程度,更加有助于挖掘知识图谱中各节点之间的潜在关系。
图2为本申请在一个实施例中提供的图谱构建过程的流程图,如图所示,包括:S101、对蕴含于信息源中的基金知识进行识别,识别所述基金知识的数据类型和数据来源;本步骤中通过对基金知识元库中的知识根据其数据类型和数据来源进行识别,例如企业内部数据库的数据为结构化数据,天天基金网等网站中的图表数据为半结构化数据,基金研报、基金经理简历、雪球社区评论等整篇文本数据为非结构化数据。S102、根据所述基金知识的数据类型和数据来源进行筛选与归纳,筛选出具有相同所述数据类型和相同所述数据来源的所述基金知识并归纳为一类;本步骤中将具有同一数据类型和同一数据来源的知识数据归纳为同一类,并且根据其不同的数据类型采取不同的抽取方法,例如对于结构化数据,通过人工设定规则来进行数据抽取, 对于半结构化数据,通过爬虫或正规表达式匹配来进行数据抽取,对于非结构化数据,通过自然语言处理来进行数据抽取。S103、根据归纳整理后的所述基金知识,建立基金知识元库。本实施例中通过对信息源中的数据进行抽取并建立了基金知识元库,为后续对所述基金知识元库中的数据进行进一步的整合提供了基础。
在一个实施例中,所述根据预设规则将所述基金知识元库中的所述基金知识进行融合,包括:对所述基金知识元库中的各实体进行ID标识;本步骤中在对所述知识元库中的知识数据根据预设融合规则进行融合之前,先对所有的实体进行ID标识,例如,基金实体和股票实体以市场交易代码作为ID标识。对所述基金知识元库中的各实体进行判断,具有统一ID标识的为同一实体,将所述同一实体进行关系和属性的合并,不具有统一ID标识的,则根据各实体属性的相似度进行合并。本步骤中对数据的融合包括新数据替换旧数据的融合,还包括根据预设融合规则对知识的质量进行评估和带权重的融合,该预设融合规则即将所述基金知识元库中的实体进行ID标识,对于同一ID标识的实体进行关系与属性的融合,对于不具有同一ID标识的实体进行相似属性的融合。
本实施例中通过对所述基金知识元库中的实体进行ID标识,再将进行ID标识的实体根据预设融合规则进行融合,将所述知识元库中的知识进行有序的整合,为后续能够快速的在所述基金知识元库中查找到需要的基金的基金信息。
在一个实施例中,所述将经过融合的所述基金知识元库存储于数据库中之后,还包括定期检查与更新所述基金知识元库中的实体、关系和属性,所述定期检查与更新包括在固定时间段内检查所述基金知 识元库中的所述实体与所述关系是否有变化,若所述实体或所述关系发生了变化,则将变化后的所述实体与所述关系更新至所述基金知识元库中。本步骤中所述数据库包括关系数据库、RDF数据库、图数据库或其中任意数据库相结合的方式,所述定期检查与更新包括在固定时间段内检查所述基金知识元库中的所述实体与所述关系是否有变化,若所述实体或所述关系发生了变化,则将变化后的所述实体与所述关系更新至所述基金知识元库中。
本实施例中通过对所述基金知识元库进行定期检查,有助于更好的维护所述基金知识元库。
图3为本申请在一个实施例中提供的图谱查询过程的流程图,如图3所示,该流程图包括:S501、接收所述查询请求,并根据所述查询请求的关键词到所述基金知识元库中匹配对应的子图;本步骤中所述后台接收所述前台发送的所述查询请求,比如用户需要查询范围内人脉最广/最核心的基金经理,所述后台获取该查询请求的关键词,比如该查询条件为复旦大学校友中人脉最广的基金经理是谁,易方达基金公司旗下最核心的基金经理是谁,并到所述基金知识元库中匹配到对应的子图。S502、对所述关键词包含的一个或多个所述子图进行节点中心度的计算;S503、根据所述查询请求的关键词到所述基金知识元库中匹配到与该关键词对应的一个或多个子图,比如匹配到的子图分别为复旦大学子图和易方达基金公司子图,并将对应子图传送至所述实时运算器中进行节点中心度的计算。S504、获取各所述子图的节点中心度的结果,并将最符合所述查询请求的所述节点中心度的结果返回至所述前台;本步骤中所述后台从所述实时运算器中获取各子土的节点中心度的计算结果,比如分别获取复旦大学子图和易方达基 金公司子图的计算结果,将最符合所述查询请求的子图的计算结果返回至所述前台。S505、所述节点中心度的结果以json数据格式返回至所述前台,并用d3js技术将所述节点中心度的结果在前台展示出路径图。
本实施例中通过对所述查询请求匹配对应子图的过程,有助于快速的定位与该查询请求最相近的图谱关系,有助于快速的对该查询请求中包含的实体进行节点中心性分析。
图4为本申请在一个实施例中提供的实时运算过程的流程图,如图4所示,该流程图包括对带权重多源最短路径的计算,所述带权重多源最短路径的计算包括根据预设规则按实体关系类型分配边的权重,再通过预设算法来计算多源最短路径;S601、所述多源最短路径包含所述子图中任意两两实体的最短路径;本步骤中根据预设规则按所述预设规则,例如基金经历之间的关系是有权重的,如果两个基金经理之间的关系若为亲属、同导师、共同管理过同一个基金的,则权重设置为1;如果两个基金经理之间的关系若为同所学校毕业的、同公司或同学的,则权重设置为2。S602、对所述子图中两两实体间的最短路径进行计算的同时,还包括点度中心度的计算,所述点度中心度包括所述子图中某个所述实体的所述关系的数量。
本实施例中通过根据预设规则计算出所述子图中两两实体间的最短路径,为后续计算所述子图中各实体的节点中心度提供基础。
在一个实施例中,接收与所述查询请求对应的子图,计算所述子图中两两实体间的最短路径,并根据所述两两实体间的最短路径来计算每个实体的节点中心度还包括:获取所述子图中任意两两实体的多源最短路径的计算结果;将所述多源最短路径的计算结果代入公式 (1)和公式(2)中分别计算对应节点的中介中心度和紧密中心度,其中公式(1)和公式(2)如下:
Figure PCTCN2018124220-appb-000001
公式(1)表示节点i在子图中的中介中心度,其中,Pjk表示任意两个节点jk之间的最短路径个数,Pjk(i)是节点jk之间通过节点i的最短路径个数,所述中介中心度包含一个节点担任其它两个节点之间最短路径的桥梁的次数,该次数越高时,所述中介节点度越大;公式(2)表示节点x在子图中的紧密中心度,d(y,x)表示节点x到任意节点y的最短路径的长度,即节点x的紧密中心度是x到其它所有节点的最短路径距离之和的倒数,所述紧密中心度包含一个节点在所述子图中所处于的网络中心位置的程度,若一个节点与许多其它节点都很靠近,该节点越接近网络中心位置;将该节点的所述中介中心度和所述紧密中心度进行求和运算,该求和结果即为节点中心度。
本实施例中通过所述子图中两两实体间的最短路径的计算结果,结合预设的算法,对所述子图中的实体进行中介中心度和紧密中心度的计算,通过将中介中心度和紧密中心度进行求和最终得到对应实体的节点中心度的计算结果,实现了多角度、动态地对全图谱和各子图中各节点的中心性分析,提高了工作效率。
基于相同的构思,本申请还提出了一种基金知识推理系统,如图5所示,所述基金知识推理系统包括抽取单元、融合单元、存储单元、输入单元、查询单元和运算单元,其中:抽取单元,设置为抽取基金知识推理平台中的信息源中的基金知识后建立基金知识元库,所述基金知识元库中包含多个子图,各子图中包括实体、关系和属性;融合单元,设置为根据预设规则将所述基金知识元库中的所述基金知 识进行融合;存储单元,设置为将经过融合的所述基金知识元库存储于数据库中;输入单元,设置为在基金知识推理平台的前台输入查询请求,并将所述查询请求发送至基金知识推理平台的后台;查询单元,设置为到所述基金知识元库中筛选出与所述查询请求对应的子图,并计算该子图中的实体的节点中心度,将所述节点中心度的计算结果返回至所述前台进行展示;运算单元,设置为接收与所述查询请求的对应的子图,计算所述子图中两两实体间的最短路径,并根据所述两两实体间的最短路径来计算每个实体的节点中心度。
在一个实施例中,所述融合单元包括:标识模块,设置为对所述基金知识元库中的各实体进行ID标识;合并模块,设置为对所述基金知识元库中的各实体进行判断,具有统一ID标识的为同一实体,将所述同一实体进行关系和属性的合并,不具有统一ID标识的,则根据各实体属性的相似度进行合并。
在一个实施例中,所述基金知识元库还包括:检查模块,设置为定期检查与更新所述基金知识元库中的实体、关系和属性,在固定时间段内检查所述基金知识元库中的所述实体与所述关系是否有变化,若所述实体或所述关系发生了变化,则将变化后的所述实体与所述关系更新至所述基金知识元库中。
在一个实施例中,所述查询单元包括:接收请求模块,设置为接收所述查询请求,并根据所述查询请求的关键词到所述基金知识元库中匹配对应的子图;获取中心度模块,设置为对所述关键词包含的一个或多个所述子图进行节点中心度的计算;返回模块,设置为获取各所述子图的节点中心度的结果,并将最符合所述查询请求的所述节点中心度的结果返回至所述前台;所述节点中心度的结果以json数据 格式返回至所述前台,并用d3js技术将所述节点中心度的结果在前台展示出路径图。
在一个实施例中,所述运算单元包括:计算模块,设置为计算带权重多源最短路径,所述带权重多源最短路径的计算包括根据预设规则按实体关系类型分配边的权重,再通过预设算法来计算多源最短路径;所述多源最短路径包含所述子图中任意两两实体的最短路径;对所述子图中两两实体间的最短路径进行计算的同时,还包括点度中心度的计算;所述点度中心度包括所述子图中某个所述实体的所述关系的数量。
在一个实施例中,所述计算模块还包括:获取路径结果,设置为获取所述子图中任意两两实体的多源最短路径的计算结果;一级计算模块,设置为将所述多源最短路径的计算结果代入公式中分别计算对应节点的中介中心度和紧密中心度;二级计算模块,设置为将该节点的所述中介中心度和所述紧密中心度进行求和运算,该求和结果即为节点中心度。
在一个实施例中,提出了一种计算机设备,所述计算机设备包括存储器和处理器,存储器中存储有计算机可读指令,计算机可读指令被处理器执行时,使得所述处理器执行上述各实施例中的所述基金知识推理方法的步骤。
在一个实施例中,提出了一种存储有计算机可读指令的存储介质,所述计算机可读指令被一个或多个处理器执行时,使得一个或多个所述处理器执行上述各实施例中的所述基金知识推理方法的步骤。其中,所述存储介质可以为非易失性存储介质。
本领域普通技术人员可以理解上述实施例的各种方法中的全部 或部分步骤是可以通过程序来指令相关的硬件来完成,该程序可以存储于一计算机可读存储介质中,存储介质可以包括:只读存储器(ROM,Read Only Memory)、随机存取存储器(RAM,Random Access Memory)、磁盘或光盘等。
以上所述实施例的各技术特征可以进行任意的组合,为使描述简洁,未对上述实施例中的各个技术特征所有可能的组合都进行描述,然而,只要这些技术特征的组合不存在矛盾,都应当认为是本说明书记载的范围。
以上所述实施例仅表达了本申请一些示例性实施例,其描述较为具体和详细,但并不能因此而理解为对本申请专利范围的限制。应当指出的是,对于本领域的普通技术人员来说,在不脱离本申请构思的前提下,还可以做出若干变形和改进,这些都属于本申请的保护范围。因此,本申请专利的保护范围应以所附权利要求为准。

Claims (20)

  1. 一种基金知识推理方法,包括:
    抽取基金知识推理平台中的信息源中的基金知识后建立基金知识元库,所述基金知识元库中包含多个子图,各子图中包括实体、关系和属性;根据预设规则将所述基金知识元库中的所述基金知识进行融合;将经过融合的所述基金知识元库存储于数据库中;在基金知识推理平台的前台输入查询请求,并将所述查询请求发送至基金知识推理平台的后台;到所述基金知识元库中筛选出与所述查询请求对应的子图,并计算该子图中的实体的节点中心度,将所述节点中心度的计算结果返回至所述前台进行展示;接收与所述查询请求对应的子图,计算所述子图中两两实体间的最短路径,并根据所述两两实体间的最短路径来计算每个实体的节点中心度。
  2. 根据权利要求1所述的一种基金知识推理方法,其中,所述抽取基金知识推理平台中的信息源中的基金知识后建立基金知识元库,包括:
    对蕴含于信息源中的基金知识进行识别,识别所述基金知识的数据类型和数据来源;根据所述基金知识的数据类型和数据来源进行筛选与归纳,筛选出具有相同所述数据类型和相同所述数据来源的所述基金知识并归纳为一类;根据归纳整理后的所述基金知识,建立基金知识元库。
  3. 根据权利要求1所述的一种基金知识推理方法,其中,所述根据预设规则将所述基金知识元库中的所述基金知识进行融合,包括:
    对所述基金知识元库中的各实体进行ID标识;对所述基金知识元库中的各实体进行判断,具有统一ID标识的为同一实体,将所述 同一实体进行关系和属性的合并,不具有统一ID标识的,则根据各实体属性的相似度进行合并。
  4. 根据权利要求1所述的一种基金知识推理方法,其中,所述将经过融合的所述基金知识元库存储于数据库中之后,还包括定期检查与更新所述基金知识元库中的实体、关系和属性,所述定期检查与更新包括在固定时间段内检查所述基金知识元库中的所述实体与所述关系是否有变化,若所述实体或所述关系发生了变化,则将变化后的所述实体与所述关系更新至所述基金知识元库中。
  5. 根据权利要求1所述的一种基金知识推理方法,其中,所述到所述基金知识元库中筛选出与所述查询请求对应的子图,并计算该子图中的实体的节点中心度,将所述节点中心度的计算结果返回至所述前台进行展示,包括:
    接收所述查询请求,并根据所述查询请求的关键词到所述基金知识元库中匹配对应的子图;对所述关键词包含的一个或多个所述子图进行节点中心度的计算;获取各所述子图的节点中心度的结果,并将最符合所述查询请求的所述节点中心度的结果返回至所述前台;所述节点中心度的结果以json数据格式返回至所述前台,并用d3js技术将所述节点中心度的结果在前台展示出路径图。
  6. 根据权利要求1所述的一种基金知识推理方法,其中,所述接收与所述查询请求对应的子图,计算所述子图中两两实体间的最短路径,并根据所述两两实体间的最短路径来计算每个实体的节点中心度,包括对带权重多源最短路径的计算,所述带权重多源最短路径的计算包括根据预设规则按实体关系类型分配边的权重,再通过预设算法来计算多源最短路径;
    所述多源最短路径包含所述子图中任意两两实体的最短路径;对所述子图中两两实体间的最短路径进行计算的同时,还包括点度中心度的计算;所述点度中心度包括所述子图中某个所述实体的所述关系的数量。
  7. 根据权利要求6所述的一种基金知识推理方法,其中,所述接收与所述查询请求对应的子图,计算所述子图中两两实体间的最短路径,并根据所述两两实体间的最短路径来计算每个实体的节点中心度还包括:
    获取所述子图中任意两两实体的多源最短路径的计算结果;将所述多源最短路径的计算结果代入公式(1)和公式(2)中分别计算应节点的中介中心度和紧密中心度,其中公式(1)如下:
    Figure PCTCN2018124220-appb-100001
    公式(1)表示节点i在子图中的中介中心度,其中,Pjk表示任意两个节点jk之间的最短路径个数,Pjk(i)是节点jk之间通过节点i的最短路径个数,所述中介中心度包含一个节点担任其它两个节点之间最短路径的桥梁的次数,该次数越高时,所述中介节点度越大,公式(2)如下:
    Figure PCTCN2018124220-appb-100002
    公式(2)表示节点x在子图中的紧密中心度,d(y,x)表示节点x到任意节点y的最短路径的长度,即节点x的紧密中心度是x到其它所有节点的最短路径距离之和的倒数,所述紧密中心度包含一个节点在所述子图中所处于的网络中心位置的程度,若一个节点与许多其它节点都很靠近,该节点越接近网络中心位置;将该节点的所述中介中 心度和所述紧密中心度进行求和运算,该求和结果即为节点中心度。
  8. 一种基金知识推理系统,包括:
    抽取单元,设置为抽取基金知识推理平台中的信息源中的基金知识后建立基金知识元库,所述基金知识元库中包含多个子图,各子图中包括实体、关系和属性;融合单元,设置为根据预设规则将所述基金知识元库中的所述基金知识进行融合;存储单元,设置为将经过融合的所述基金知识元库存储于数据库中;输入单元,设置为在基金知识推理平台的前台输入查询请求,并将所述查询请求发送至基金知识推理平台的后台;查询单元,设置为到所述基金知识元库中筛选出与所述查询请求对应的子图,并计算该子图中的实体的节点中心度,将所述节点中心度的计算结果返回至所述前台进行展示;运算单元,设置为接收与所述查询请求的对应的子图,计算所述子图中两两实体间的最短路径,并根据所述两两实体间的最短路径来计算每个实体的节点中心度。
  9. 根据权利要求8所述的基金知识推理系统,其中,所述抽取单元包括:
    识别模块,设置为对蕴含于信息源中的基金知识进行识别,识别所述基金知识的数据类型和数据来源;筛选模块,设置为根据所述基金知识的数据类型和数据来源进行筛选与归纳,筛选出具有相同所述数据类型和相同所述数据来源的所述基金知识并归纳为一类;建立模块,设置为根据归纳整理后的所述基金知识,建立基金知识元库。
  10. 根据权利要求8所述的基金知识推理系统,其中,所述融合单元包括:
    标识模块,设置为对所述基金知识元库中的各实体进行ID标识; 合并模块,设置为对所述基金知识元库中的各实体进行判断,具有统一ID标识的为同一实体,将所述同一实体进行关系和属性的合并,不具有统一ID标识的,则根据各实体属性的相似度进行合并。
  11. 根据权利要求8所述的基金知识推理系统,其中,所述基金知识元库还包括:检查模块,设置为定期检查与更新所述基金知识元库中的实体、关系和属性。
  12. 根据权利要求11所述的基金知识推理系统,其中,所述检查模块包括:
    定期检查模块,设置为在固定时间段内检查所述基金知识元库中的所述实体与所述关系是否有变化,若所述实体或所述关系发生了变化,则将变化后的所述实体与所述关系更新至所述基金知识元库中。
  13. 根据权利要求8所述的基金知识推理系统,其中,所述查询单元包括:
    接收请求模块,设置为接收所述查询请求,并根据所述查询请求的关键词到所述基金知识元库中匹配对应的子图;获取中心度模块,设置为对所述关键词包含的一个或多个所述子图进行节点中心度的计算;返回模块,设置为获取各所述子图的节点中心度的结果,并将最符合所述查询请求的所述节点中心度的结果返回至所述前台;所述节点中心度的结果以json数据格式返回至所述前台,并用d3js技术将所述节点中心度的结果在前台展示出路径图。
  14. 根据权利要求8所述的基金知识推理系统,其中,所述运算单元包括:
    计算模块,设置为计算带权重多源最短路径,所述带权重多源最短路径的计算包括根据预设规则按实体关系类型分配边的权重,再通 过预设算法来计算多源最短路径;所述多源最短路径包含所述子图中任意两两实体的最短路径;对所述子图中两两实体间的最短路径进行计算的同时,还包括点度中心度的计算;所述点度中心度包括所述子图中某个所述实体的所述关系的数量。
  15. 根据权利要求14所述的基金知识推理系统,其中,所述计算模块还包括:
    获取路径模块,设置为获取所述子图中任意两两实体的多源最短路径的计算结果;一级计算模块,设置为将所述多源最短路径的计算结果代入公式(1)和公式(2)中分别计算对应节点的中介中心度和紧密中心度,其中公式如下:
    Figure PCTCN2018124220-appb-100003
    公式(1)表示节点i在子图中的中介中心度,其中,Pjk表示任意两个节点jk之间的最短路径个数,Pjk(i)是节点jk之间通过节点i的最短路径个数,所述中介中心度包含一个节点担任其它两个节点之间最短路径的桥梁的次数,该次数越高时,所述中介节点度越大,公式如下:公式(2)表示节点x在子图中的紧密中心度,d(y,x)表示节点x到任意节点y的最短路径的长度,即节点x的紧密中心度是x到其它所有节点的最短路径距离之和的倒数,所述紧密中心度包含一个节点在所述子图中所处于的网络中心位置的程度,若一个节点与许多其它节点都很靠近,该节点越接近网络中心位置;二级计算模块,设置为将该节点的所述中介中心度和所述紧密中心度进行求和运算,该求和结果即为节点中心度。
  16. 一种计算机设备,包括存储器和处理器,所述存储器中存储有计算机可读指令,所述计算机可读指令被所述处理器执行时,使得 所述处理器执行以下步骤:抽取基金知识推理平台中的信息源中的基金知识后建立基金知识元库,所述基金知识元库中包含多个子图,各子图中包括实体、关系和属性;根据预设规则将所述基金知识元库中的所述基金知识进行融合;将经过融合的所述基金知识元库存储于数据库中;在基金知识推理平台的前台输入查询请求,并将所述查询请求发送至基金知识推理平台的后台;到所述基金知识元库中筛选出与所述查询请求对应的子图,并计算该子图中的实体的节点中心度,将所述节点中心度的计算结果返回至所述前台进行展示;接收与所述查询请求对应的子图,计算所述子图中两两实体间的最短路径,并根据所述两两实体间的最短路径来计算每个实体的节点中心度。
  17. 根据权利要求16所述的一种计算机设备,其中,所述抽取基金知识推理平台中的信息源中的基金知识后建立基金知识元库时,使得所述处理器执行以下步骤:
    对蕴含于信息源中的基金知识进行识别,识别所述基金知识的数据类型和数据来源;根据所述基金知识的数据类型和数据来源进行筛选与归纳,筛选出具有相同所述数据类型和相同所述数据来源的所述基金知识并归纳为一类;根据归纳整理后的所述基金知识,建立基金知识元库。
  18. 根据权利要求16所述的一种计算机设备,其中,所述根据预设规则将所述基金知识元库中的所述基金知识进行融合时,使得所述处理执行器执行以下步骤:
    对所述基金知识元库中的各实体进行ID标识;对所述基金知识元库中的各实体进行判断,具有统一ID标识的为同一实体,将所述同一实体进行关系和属性的合并,不具有统一ID标识的,则根据各 实体属性的相似度进行合并。
  19. 一种存储有计算机可读指令的存储介质,所述计算机可读指令被一个或多个处理器执行时,使得一个或多个所述处理器执行以下步骤:抽取基金知识推理平台中的信息源中的基金知识后建立基金知识元库,所述基金知识元库中包含多个子图,各子图中包括实体、关系和属性;根据预设规则将所述基金知识元库中的所述基金知识进行融合;将经过融合的所述基金知识元库存储于数据库中;在基金知识推理平台的前台输入查询请求,并将所述查询请求发送至基金知识推理平台的后台;到所述基金知识元库中筛选出与所述查询请求对应的子图,并计算该子图中的实体的节点中心度,将所述节点中心度的计算结果返回至所述前台进行展示;接收与所述查询请求对应的子图,计算所述子图中两两实体间的最短路径,并根据所述两两实体间的最短路径来计算每个实体的节点中心度。
  20. 根据权利要求19所述的一种存储有计算机可读指令的存储介质,其中,所述抽取基金知识推理平台中的信息源中的基金知识后建立基金知识元库时,使得所述一个或多个处理器执行以下步骤:
    对蕴含于信息源中的基金知识进行识别,识别所述基金知识的数据类型和数据来源;根据所述基金知识的数据类型和数据来源进行筛选与归纳,筛选出具有相同所述数据类型和相同所述数据来源的所述基金知识并归纳为一类;根据归纳整理后的所述基金知识,建立基金知识元库。
PCT/CN2018/124220 2018-09-03 2018-12-27 基金知识推理方法、系统、计算机设备和存储介质 Ceased WO2020048058A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811019068.9 2018-09-03
CN201811019068.9A CN109299090B (zh) 2018-09-03 2018-09-03 基金中心度计算方法、系统、计算机设备和存储介质

Publications (1)

Publication Number Publication Date
WO2020048058A1 true WO2020048058A1 (zh) 2020-03-12

Family

ID=65166045

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/124220 Ceased WO2020048058A1 (zh) 2018-09-03 2018-12-27 基金知识推理方法、系统、计算机设备和存储介质

Country Status (2)

Country Link
CN (1) CN109299090B (zh)
WO (1) WO2020048058A1 (zh)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111813951A (zh) * 2020-06-18 2020-10-23 国网上海市电力公司 一种基于技术图谱的关键点识别方法
CN112445875A (zh) * 2020-12-14 2021-03-05 招商局金融科技有限公司 数据关联及检验方法、装置、电子设备及存储介质
CN113362113A (zh) * 2021-06-07 2021-09-07 中国工商银行股份有限公司 一种企业行为偏好的获取方法及装置
CN113723047A (zh) * 2021-07-27 2021-11-30 山东旗帜信息有限公司 一种基于法律文件的图谱构建方法、设备及介质
CN113886594A (zh) * 2021-09-14 2022-01-04 浙江嘉兴数字城市实验室有限公司 基于知识图谱的社会治理领域要素的深度挖掘方法
CN114238522A (zh) * 2021-09-30 2022-03-25 武汉众智数字技术有限公司 一种基于知识图谱的家族关系推理方法及系统

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111814001B (zh) * 2019-04-11 2024-05-10 杭州海康威视数字技术股份有限公司 反馈信息的方法和装置
CN110347757A (zh) * 2019-07-12 2019-10-18 北京熵简科技有限公司 一种智能投研知识图谱系统
CN111369331A (zh) * 2020-02-28 2020-07-03 中国工商银行股份有限公司 交叉性风险防控方法及装置
CN111782820B (zh) * 2020-06-30 2024-05-17 京东科技控股股份有限公司 知识图谱创建方法及装置、可读存储介质及电子设备
CN112163094B (zh) * 2020-08-25 2022-10-14 中国科学院计算机网络信息中心 一种科技资源汇聚与持续服务方法及装置
CN112347113B (zh) * 2020-09-16 2021-12-14 北京中兵数字科技集团有限公司 航空数据融合方法、航空数据融合装置和存储介质
CN112199960B (zh) * 2020-11-12 2021-05-25 北京三维天地科技股份有限公司 一种标准知识元粒度解析系统
CN112948469B (zh) * 2021-04-16 2023-10-13 平安科技(深圳)有限公司 数据挖掘方法、装置、计算机设备及存储介质
CN115186099A (zh) * 2022-05-31 2022-10-14 民生科技有限责任公司 一种基于多维度构建知识图谱的对公营销方法和系统
CN115994231B (zh) * 2023-03-24 2023-05-16 西南石油大学 稠油配汽优化的知识图谱优化方法
CN116595485A (zh) * 2023-05-25 2023-08-15 江苏省联合征信有限公司 一种图数据集成计算方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105389622A (zh) * 2015-10-20 2016-03-09 西安交通大学 一种基于知识地图的多约束学习路径推荐方法
CN105765576A (zh) * 2013-11-15 2016-07-13 微软技术许可有限责任公司 使用数据结构处理搜索查询
CN106294588A (zh) * 2016-07-28 2017-01-04 广东中标数据科技股份有限公司 快速搜索所要查询内容的方法及装置

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103179052B (zh) * 2011-12-20 2016-03-02 中国科学院声学研究所 一种基于接近度中心性的虚拟资源分配方法及系统
CN104636424A (zh) * 2014-12-02 2015-05-20 南昌大学 一种基于图谱分析构建文献综述框架的方法
CN106919689B (zh) * 2017-03-03 2018-05-11 中国科学技术信息研究所 基于术语释义知识单元的专业领域知识图谱动态构建方法
CN107103403A (zh) * 2017-03-06 2017-08-29 广东工业大学 一种用cio/cto社交网络预测科技公司业绩的方法
CN108052547B (zh) * 2017-11-27 2019-09-27 华中科技大学 基于问句和知识图结构分析的自然语言问答方法及系统

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105765576A (zh) * 2013-11-15 2016-07-13 微软技术许可有限责任公司 使用数据结构处理搜索查询
CN105389622A (zh) * 2015-10-20 2016-03-09 西安交通大学 一种基于知识地图的多约束学习路径推荐方法
CN106294588A (zh) * 2016-07-28 2017-01-04 广东中标数据科技股份有限公司 快速搜索所要查询内容的方法及装置

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111813951A (zh) * 2020-06-18 2020-10-23 国网上海市电力公司 一种基于技术图谱的关键点识别方法
CN112445875A (zh) * 2020-12-14 2021-03-05 招商局金融科技有限公司 数据关联及检验方法、装置、电子设备及存储介质
CN112445875B (zh) * 2020-12-14 2024-03-01 招商局金融科技有限公司 数据关联及检验方法、装置、电子设备及存储介质
CN113362113A (zh) * 2021-06-07 2021-09-07 中国工商银行股份有限公司 一种企业行为偏好的获取方法及装置
CN113723047A (zh) * 2021-07-27 2021-11-30 山东旗帜信息有限公司 一种基于法律文件的图谱构建方法、设备及介质
CN113886594A (zh) * 2021-09-14 2022-01-04 浙江嘉兴数字城市实验室有限公司 基于知识图谱的社会治理领域要素的深度挖掘方法
CN114238522A (zh) * 2021-09-30 2022-03-25 武汉众智数字技术有限公司 一种基于知识图谱的家族关系推理方法及系统

Also Published As

Publication number Publication date
CN109299090A (zh) 2019-02-01
CN109299090B (zh) 2023-05-30

Similar Documents

Publication Publication Date Title
WO2020048058A1 (zh) 基金知识推理方法、系统、计算机设备和存储介质
CN109492111B (zh) 最短路径查询方法、系统、计算机设备和存储介质
US11985037B2 (en) Systems and methods for conducting more reliable assessments with connectivity statistics
CN112085205B (zh) 用于自动训练机器学习模型的方法和系统
US10475045B2 (en) Database management methods, systems, and devices for identifying related customer support tickets
US9922134B2 (en) Assessing and scoring people, businesses, places, things, and brands
US8874616B1 (en) Method and apparatus for fusion of multi-modal interaction data
JP4920023B2 (ja) オブジェクト間競合指標計算方法およびシステム
US10311106B2 (en) Social graph visualization and user interface
CN111309822B (zh) 用户身份识别方法及装置
US11423036B2 (en) Systems and methods for selecting datasets
EP3881197A1 (en) Systems and method for scoring entities and networks in a knowledge graph
CN111931069B (zh) 用户兴趣确定方法、装置及计算机设备
CN110020433A (zh) 一种基于企业关联关系的工商高管人名消歧方法
US11373101B2 (en) Document analyzer
CN109408643B (zh) 基金相似度计算方法、系统、计算机设备和存储介质
KR20160104064A (ko) 복잡한 양자 또는 다자 상대방 관계를 탐색하기 위해 이용되는 다차원 재귀적 학습 과정 및 시스템
CN120031113A (zh) 一种基于时序知识图谱的数据处理方法及系统
EP3652660B1 (en) Systems and methods for joining datasets
EP2731021A1 (en) Apparatus, program, and method for reconciliation processing in a graph database
CN104199924B (zh) 选择具有快照关系的网络表格的方法及装置
JP7607537B2 (ja) ユーザに対して因果ループ図を表示するシステム及び方法
Kpiebaareh et al. User-connection behaviour analysis in service management using bipartite labelled property graph
Jiang et al. Optimizing urban commercial landscapes with poi strategic group analysis: A meta-learning hypergraph approach
CN120763259B (zh) 一种基于机器学习的数据分析方法及系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18932800

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18932800

Country of ref document: EP

Kind code of ref document: A1