A relationship path retrieval method and system for a social resource map

Abstract

This invention discloses a method and system for retrieving relationship paths in a network resource map, comprising a processor, a memory, and a data access unit, a heterogeneous graph construction unit, a high-connectivity organization identification unit, an organization compression index unit, a resource candidate retrieval unit, a path search unit, a virtual bridging processing unit, a pruning and sorting unit, and a path output unit. This method and system for retrieving relationship paths in a network resource map, by constructing a heterogeneous network resource graph, identifying high-connectivity organization nodes, establishing an organization member partition compression index, determining a candidate set based on a resource inverted index, and controlling the expansion of high-connectivity organization nodes through virtual bridging edges during the path search process, can reduce the computational overhead in large-scale organization member relationship retrieval while ensuring the reachability and interpretability of network resource paths, and improve the effectiveness of relationship path retrieval results.

Description

Technical Field

[0001] This invention relates to the field of information processing technology, specifically to a method and system for retrieving relationship paths from a network of personal resources map. Background Technology

[0002] In networking platforms, existing technologies typically establish a network of relationships through user address books, electronic business cards, information on chambers of commerce and associations, resource tags, and cooperation records. Based on keyword search, friend relationship expansion, or shortest path search, they return possible contacts or resource providers to users. However, when connecting a large number of members to organizational nodes such as chambers of commerce, associations, colleges, and communities, directly traversing these organizational nodes as ordinary nodes can easily lead to drawbacks such as a sharp increase in the number of relationship edges, an excessively large path search range, longer query response time, and duplicated and uninterpretable returned paths.

[0003] Several domestic and international patents have improved the calculation of personal connections and the display of social graph paths. For example, patent publication number CN104239423A describes a method and system for obtaining personal connection distances. This method expands direct or indirect contacts by progressively expanding the phone numbers in the mobile terminal's address book, and calculates the personal connection distance between the target contact and the user by comprehensively considering the personal connection level, call frequency, and call duration. Another example is patent publication number US20130254305A1, "Techniques for identifying and presenting connection paths," which identifies the connection path between two entities through a social graph containing entity nodes such as people, companies, educational institutions, and groups, and related edges. It then aggregates the path scores based on edge scores and displays the optimal connection path. Therefore, existing technologies have already solved some of the problems in personal connection retrieval in terms of relationship level calculation, social graph modeling, path identification, and path display.

[0004] However, the aforementioned existing patents and related technologies still mainly focus on the distance of contact relationships, the display of social graph connection paths, or general path scores. They do not address the issue of establishing a partitioned and compressed index of organizational members for highly connected organizational nodes that include a large number of chambers of commerce, associations, colleges, and communities. Instead, they do not propose a mechanism for controlled expansion and path pruning when highly connected organizational nodes are found in the path search. Therefore, problems still exist, such as path explosion caused by the full expansion of organizational members, the separation between resource retrieval and relationship path retrieval, high noise in weakly connected organizational paths, and difficulty in maintaining path validity after changes in organizational members or resource tags. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for retrieving relationship paths based on a network of personal resources, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method and system for retrieving relationship paths based on a network resource map, comprising: S1. Acquire user, business card, organization, resource tags and cooperation event data to construct a heterogeneous network resource map; S2. Identify high-connectivity organizational nodes based on the number of organizational nodes' members, the number of connected edges, or historical expansion overhead, and build an organizational member partitioned compressed index; S3. Receive a path retrieval request containing the starting user and target resource conditions, and determine candidate resource nodes, candidate user nodes and candidate organization partitions through the resource inverted index; S4. Perform a path search with a limited number of hops starting from the user. For ordinary nodes, expand the edge based on the effective relationship. For high-connection organization nodes, generate virtual bridge edges based on the hit candidate organization partitions, without fully expanding the members of the high-connection organization node. S5. When the virtual bridge edge meets the resource matching condition and the expansion budget condition, select a candidate member from the corresponding organization member partition as the next hop node to continue the search; S6. Prune and sort the candidate paths, and output the network resource relationship path from the starting user to the target resource party.

[0007] Further elaboration reveals that the nodes in the heterogeneous network resource graph include user nodes, business card nodes, organization nodes, resource tag nodes, and cooperation event nodes. The relationship edges in the heterogeneous network resource graph include at least two of the following: business card exchange edges, organization affiliation edges, resource supply edges, cooperation proof edges, and recommendation relationship edges. Each relationship edge is configured with at least one of the following attributes: edge type, edge status, relationship strength, proof level, and update version.

[0008] Further, when identifying the highly connected organizational node, the number of members, the number of connection edges, and the number of times or time spent expanding members in historical queries are counted. When any statistical result reaches the corresponding preset threshold, the organizational node is marked as a highly connected organizational node, and the highly connected organizational node is used as an organizational node that needs to undergo organizational member partitioning and compression indexing.

[0009] Further elaborated, the organizational member partition compression index partitions members under the same highly connected organizational node according to at least two of the following dimensions: region, industry, role, resource tag, member status, active time, and relationship proof. Each organizational member partition records a partition identifier, the organization identifier, the number of members in the partition, a partition member index list, a candidate member list, a resource tag set, and a partition version number.

[0010] Further, determining candidate resource nodes, candidate personnel nodes, and candidate organizational partitions through the resource inverted index includes: converting the target resource conditions into standard resource tags or resource categories, and using the standard resource tags or resource categories in combination with at least one of regional constraints and industry constraints as search conditions to query resource tag nodes that match the target resource conditions, personnel nodes that are directly associated with the target resource conditions, and organizational member partitions containing matching members.

[0011] Further elaborated, the virtual bridging edge is a logical edge temporarily generated during the path search process based on the highly connected organization node and its matched candidate organization partitions. The virtual bridging edge is used to represent the path status of continuing to search for candidate members from the highly connected organization node through the corresponding organization member partition. The virtual bridging edge records the organization identifier, partition identifier, number of candidate members, partition matching degree, organization bridging cost, and the upper limit of the number of members allowed to be expanded.

[0012] Further elaboration, the virtual bridge edge satisfies the resource matching conditions and expansion budget conditions as follows: the matching degree between the corresponding organization member partition and the target resource condition reaches a preset matching threshold, the current path hop count does not exceed the limited hop count, the organization expansion budget, partition expansion budget and member expansion budget corresponding to the current path retrieval request are not exhausted, and there are candidate members in the corresponding organization member partition that are valid and associated with the target resource condition.

[0013] Further, pruning candidate paths includes: pruning candidate paths that exceed the maximum hop count limit; pruning candidate paths containing invalid relation edges; pruning candidate paths that repeatedly pass through the same node; pruning candidate paths that do not conform to the preset edge type sequence; and pruning candidate paths that reach the same target resource and are dominated by other candidate paths in terms of path hop count, organizational bridging cost, resource matching degree, and relation strength.

[0014] Further, the ranking of candidate paths includes: determining the ranking result of candidate paths based on resource matching degree, relationship strength, relationship proof status, path hop count, and organizational bridging cost; when outputting the network resource relationship path, at least the path node sequence, relationship edge type, traversed organizational information, target resource tag, path hop count, and relationship proof summary are output.

[0015] A relationship path retrieval system for a network resource map includes a processor, a memory, and a data access unit, a heterogeneous graph construction unit, a high-connectivity organization identification unit, an organization compression index unit, a resource candidate retrieval unit, a path search unit, a virtual bridging processing unit, a pruning and sorting unit, and a path output unit. The heterogeneous graph construction unit constructs a heterogeneous network resource map containing users, business cards, organizations, resource tags, and collaborative events. The high-connectivity organization identification unit and the organization compression index unit identify high-connectivity organization nodes and establish a compressed index for organization member partitions. The resource candidate retrieval unit obtains candidate resource nodes, candidate member nodes, and candidate organization partitions based on target resource conditions. The path search unit and the virtual bridging processing unit generate virtual bridging edges and controllably expand candidate members when a high-connectivity organization node is found during path searching. The pruning and sorting unit and the path output unit prune and sort the candidate paths and output the network resource relationship path.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This relationship path retrieval method and system for a network resource map, by constructing a heterogeneous network resource map, identifying highly connected organizational nodes, establishing a partitioned compressed index for organizational members, determining the candidate set based on the resource inverted index, and controlling the expansion of highly connected organizational nodes through virtual bridging edges during the path search process, can reduce the computational overhead in the large-scale organizational member relationship retrieval process and improve the effectiveness of the relationship path retrieval results while ensuring the accessibility and interpretability of network resource paths.

[0017] 1. After identifying highly connected organizational nodes, instead of directly expanding all members under that organizational node, a partitioned compressed index of organizational members is built, and controlled member reading is performed only after a candidate organizational partition is hit. This reduces the path explosion problem caused by highly connected organizational nodes such as chambers of commerce, associations, colleges, and communities, and reduces the number of nodes expanded and the burden of traversing relationship edges in a single search.

[0018] 2. Before performing a path search with a limited number of hops, candidate resource nodes, candidate member nodes, and candidate organization partitions are determined through the resource inverted index. This ensures that the path search process revolves around the target resource conditions. When a highly connected organization node is found, a virtual bridging edge is used to connect the hit organization member partition, thereby reducing the participation of irrelevant contacts and irrelevant organization members in the search and improving the targeting of the network resource path search.

[0019] 3. Candidate paths are pruned based on node duplication, edge status, edge type sequence, maximum hop count, and path dominance. They are then sorted by resource matching degree, relationship strength, relationship proof status, and organizational bridging cost. Path validity can be maintained by using version information of organizational member partitions, thereby reducing duplicate and weakly connected paths and improving the interpretability and stability of returned relationship paths. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the system structure of the present invention; Figure 2 This is a schematic diagram of the heterogeneous network resource graph structure of the present invention; Figure 3 This is a flowchart of the high-connectivity organization node identification process of the present invention; Figure 4 This is a schematic diagram of the organizational member partition compression index of the present invention; Figure 5 This is a flowchart of the relationship path retrieval process of the present invention; Figure 6 This is a flowchart of the virtual bridging edge generation and controlled unfolding process of the present invention; Figure 7 This is a flowchart of the path pruning and sorting process of the present invention; Figure 8 This is a schematic diagram illustrating the incremental update of the partition index in this invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figures 1-8 The present invention provides the following technical solution: A relationship path retrieval system for a network resource map includes a processor, a memory, and a data access unit stored in the memory and accessed by the processor, a heterogeneous graph construction unit, a high-connectivity organization identification unit, an organization compression indexing unit, a resource candidate retrieval unit, a path search unit, a virtual bridging processing unit, a pruning and sorting unit, and a path output unit. The heterogeneous graph construction unit is used to construct a heterogeneous network resource map containing users, business cards, organizations, resource tags, and cooperation events. The high-connectivity organization identification unit and the organization compression indexing unit are used to identify high-connectivity organization nodes and establish a compressed index of organization member partitions. The resource candidate retrieval unit is used to obtain candidate resource nodes, candidate member nodes, and candidate organization partitions based on target resource conditions. The path search unit and the virtual bridging processing unit are used to generate virtual bridging edges and expand candidate members in a controlled manner when a high-connectivity organization node is found in the path search. The pruning and sorting unit and the path output unit are used to prune and sort the candidate paths and output the network resource relationship path.

[0023] The retrieval method is as follows: S1. In this embodiment, the server obtains user, business card, organization, resource tag and cooperation event data from the mini-program user registration information, electronic business card information, organization member information, resource tag information and cooperation event records. The user data includes at least user identifier, region, industry, identity status and activity status. The business card data includes at least business card identifier, company name, job information and business card status. The organization data includes at least organization identifier, organization type, region, industry scope, number of members and organization status. The resource tag data includes at least resource type, keywords, industry, region and validity period. The cooperation event data includes at least event type, participants, occurrence time and proof status.

[0024] The acquisition of user, business card, organization, resource tag, and cooperation event data does not directly read the content displayed on the front-end page. Instead, it reads structured data from the back-end business database, business card exchange records, chamber of commerce or association member tables, resource tag tables, and cooperation or activity record tables. User identifiers, business card identifiers, organization identifiers, resource tag identifiers, and event identifiers from different sources are uniformly encoded so that the same object has a unique node identifier in the subsequent graph construction process.

[0025] After data acquisition, the system performs validity processing on user, business card, organization, resource tag, and cooperation event data. Specifically, it filters data records that are deleted, expired, or have exceeded their validity period, retains data records that are valid or pending confirmation but can be used for retrieval, and standardizes fields such as region, industry, role, and resource tag so that fields with the same meaning but different expressions from different sources can be mapped to the same standard field.

[0026] In this embodiment, constructing a heterogeneous network resource graph refers to converting different types of network resource objects into different types of nodes in a graph data structure, and recording the business connection relationships between nodes through relational edges. Specifically, user data is converted into user nodes, business card data is converted into business card nodes, organization data is converted into organization nodes, resource tag data is converted into resource tag nodes, and cooperation event data is converted into cooperation event nodes.

[0027] When constructing a heterogeneous network resource graph, the system generates business card exchange edges based on business card exchange records, organization affiliation edges based on organization member records, resource supply edges based on the correspondence between users or business cards and resource tags, cooperation proof edges based on the participants in cooperation events, and recommendation relationship edges based on referral relationships or historical referral records. Each relationship edge records the starting node, ending node, edge type, edge status, relationship strength, proof level, visibility status, and latest version.

[0028] To avoid creating a large number of unnecessary pairwise connections between members during the initial graph construction phase, this embodiment does not directly interconnect all members within the same chamber of commerce, association, college, or community when constructing the heterogeneous network resource graph. Instead, it retains the organizational affiliation edges between member nodes and organizational nodes, making the organizational nodes the foundational nodes for subsequent identification of highly connected organizational nodes and the establishment of organizational member partitioning and compression indexes.

[0029] In this embodiment, the network resource map refers to a data set that represents the relationship between users, business cards, organizations, resource tags and cooperation events using a graph data structure. The heterogeneous network resource map refers to a graph structure that contains different node types and different relationship edge types. A relationship edge is a connection record between two nodes that can be identified and retrieved by the system. The definition is used to enable the data after constructing the heterogeneous network resource map to be jointly called by subsequent path search, organization compression index and candidate path pruning.

[0030] Through the above processing, the execution result of S1 is to generate a heterogeneous network resource graph containing user nodes, business card nodes, organization nodes, resource tag nodes, cooperation event nodes and corresponding relationship edges. The heterogeneous network resource graph not only preserves the interpretable connection relationship between network resource objects, but also avoids the full expansion of highly connected organization members in the data construction stage, providing a data foundation for subsequent identification of highly connected organization nodes according to the connection status of organization nodes.

[0031] S2. In this embodiment, after the system completes the construction of the heterogeneous network resource graph, it identifies high-connectivity organizational nodes based on the connection scale and historical retrieval load of organizational nodes in the graph structure, according to the number of members, the number of connection edges, or the historical expansion overhead of the organizational nodes, and establishes a partitioned compressed index of organizational members. The organizational nodes include chamber of commerce nodes, association nodes, college nodes, and community nodes. High-connectivity organizational nodes refer to organizational nodes that are prone to triggering a large number of member expansions and causing an increase in query load during path search.

[0032] When the system identifies highly connected organizational nodes based on the number of members, the number of connection edges, or the historical expansion cost, it first reads the number of user nodes and business card nodes attached to each organizational node from the organizational member table. Then, it reads the number of organizational affiliation edges, resource supply edges, cooperation proof edges, and recommendation relationship edges corresponding to the organizational node from the heterogeneous network resource graph. Finally, it reads the number of expansion nodes, expansion time, and trigger count generated by the organizational node in historical queries from the path retrieval log.

[0033] The table below shows the high-connectivity organization node identification parameters, which are used to explain the data sources and processing results used by the system to determine whether an organization node needs to enter the partition compression process.

[0034] Parameter name Data source Judgment meaning Processing results Number of members Organization Member Table Determine if there are too many members attached to the organization. Nodes that reach a threshold are marked as high-connectivity organization nodes. Number of connected edges Heterogeneous network resource map Determine if an organization node is connected to a large number of relation edges. Normal full expansion is disabled when the threshold is reached. Historical unfolding expenses Path retrieval log Determine if the organization nodes are causing excessive query load. When the threshold is reached, the process switches to partitioned compression. Organizational status Organization Information Form Determine if an organization is valid or eligible for retrieval. Invalid organizations do not establish searchable partitions. Updated version Organizational update records Determine if organizational member or resource data has changed. The change triggers a partition index update. When the number of members of a certain organization node reaches a preset threshold, or the number of connected edges reaches a preset threshold, or the historical expansion cost reaches a preset expansion cost threshold, the system identifies the organization node as a high-connectivity organization node and writes the high-connectivity flag, trigger type, and judgment time into the node attributes of the organization node so that subsequent path search can distinguish between ordinary organization nodes and high-connectivity organization nodes.

[0035] For the identified highly connected organizational nodes, the system performs the creation of an organizational member partition compressed index. Specifically, it reads the member records under the highly connected organizational node that are valid or eligible for retrieval, and groups them according to at least two of the following dimensions: region, industry, role, resource tag, member status, active time, and relationship proof, to form organizational member partitions that can be directly matched by resource retrieval conditions.

[0036] In this embodiment, the region is used to represent the city, district, or business coverage area of ​​the member; the industry is used to represent the industry or service industry to which the member belongs; the role is used to represent the identity category of the member in the organization; the resource tag is used to represent the type of resources that the member can provide; the member status is used to represent whether the member is valid, expired, or pending confirmation; the active time is used to represent whether the member has recently participated in platform activities or relationship updates; and the relationship proof is used to represent whether the member has exchanged business cards, participated in collaborative events, or had a record of being at the same event.

[0037] The table below is a compressed index table for organization member partitions, which is used to explain the data content and subsequent uses of each organization member partition after the compressed index for organization member partitions is created.

[0038] Field Name Technical meaning Subsequent uses Partition identifier Uniquely represents an organization member partition Referenced as candidate organizational partitions and virtual bridging edges Organizational identifier This indicates the high-connectivity organization node corresponding to this partition. Organizational nodes used to trace the relationship path Number of members in the partition Indicates the size of the members contained in the partition. Used to determine the cost of partition expansion Partition Member Index List Record member nodes or name card nodes within this partition. Used to read candidate members Candidate Member List Member set pre-sorted by activity level, relationship proof status, and resource matching degree Priority reading during controlled deployment Resource tag collection This indicates the resource tags associated with members within a partition. Used for matching with target resource conditions. Partition version number Indicates the current update status of the data in this partition. Used for cache validation and incremental updates When creating a compressed index for organizational member partitions, the system does not change the original organizational affiliation between organizational nodes and member nodes in the heterogeneous network resource graph. Instead, it creates a partition index layer that can be accessed by the search engine outside of highly connected organizational nodes. This allows the query process to first determine whether a certain organizational member partition is related to the target resource conditions before deciding whether to read the specific members in that partition.

[0039] For example, for a chamber of commerce organization node with a large number of members, the system can group members that meet the criteria of Shenzhen region, enterprise service industry, financing resource tags, and valid member status into the same organization member partition, and record the candidate member list and resource tag set in the organization member partition. When users search for enterprise financing resources later, the system will first target the organization member partition instead of directly traversing all members under the chamber of commerce.

[0040] Through the above processing, S2 obtains a set of highly connected organization nodes and a corresponding compressed index of organization member partitions. The set of highly connected organization nodes is used to control the expansion method of organization nodes in subsequent path search, and the compressed index of organization member partitions is used to support candidate organization partition hits, virtual bridge edge generation, and controlled expansion of candidate members.

[0041] S3. In this embodiment, after the system completes the identification of high-connectivity organization nodes and the establishment of the organization member partition compression index, it receives a path retrieval request containing the starting user and target resource conditions. The path retrieval request is submitted by the user at the network resource retrieval portal, and carries at least the starting user identifier and target resource keywords, and may carry constraint fields such as regional constraints, industry constraints, maximum number of relationship hops, number of return paths, and whether relationship proof is required.

[0042] Receiving a path retrieval request containing the starting user and target resource conditions does not simply involve receiving the natural language text input from the front end. Instead, it encapsulates the front end input, the currently logged-in user identifier, the user-selected region conditions, industry conditions, and return quantity conditions into a retrieval request object that the back end can process, enabling subsequent searches to be constrained around the starting user and target resource conditions.

[0043] The following table is the path retrieval request field table, which describes the main fields that the system receives in S3 and participates in subsequent candidate retrieval.

[0044] Field Name Technical meaning Processing purpose Starting point user ID User node identifier that initiated the path search Determine the starting node for subsequent path search Target resource keywords User-inputted or selected resource description Convert to standard resource tags or resource categories Regional constraints User-defined city, district / county, or business coverage area Filtering out candidates that do not meet the region criteria Industry constraints User-defined industry categories or service industries Filter out candidates that do not meet industry criteria. Maximum number of relationship hops Maximum number of connection levels allowed in this search Passed to subsequent path search with limited jump count Number of returned paths The expected number of relation paths to be returned in this search Output size after controlling candidate path sorting Relationship proof requirements Do you require the exchange of business cards, collaborations, or recording of events? Retention conditions affecting candidates and candidate paths After receiving a path retrieval request, the system parses the target resource conditions and converts them into standard resource tags or resource categories. The target resource keywords can be matched with resource tag thesaurus, industry classification table, and synonym mapping table. For example, input content such as corporate financing, financing services, equity financing, and bank loans can be merged into a preset set of financing-related resource tags.

[0045] After the target resource conditions are standardized, the system uses the resource inverted index to determine candidate resource nodes, candidate personnel nodes, and candidate organizational partitions. The resource inverted index uses at least one of the standard resource tags, resource categories, regional constraints, and industry constraints as the index key and records the data structure of the corresponding resource nodes, personnel nodes, and organizational member partitions.

[0046] The table below shows the data structure of the resource inverted index, which illustrates how the resource inverted index supports the pre-positioning of candidate objects.

[0047] Index key type Corresponding index content generated candidate objects Standard resource tags Resource tag nodes, personnel nodes, and organization member partitions associated with this resource tag Candidate resource nodes, candidate personnel nodes, candidate organization partitions Resource Categories The collection of objects associated with the tags under this resource category Candidate resource nodes and candidate member nodes Regional constraints Personnel nodes and organization member partitions located in or covering the region Candidate member nodes and candidate organization partitions Industry constraints Personnel nodes and organizational members belonging to or serving this industry Candidate member nodes and candidate organization partitions Tag and region combination A set of objects that simultaneously meet resource tag and region criteria Simplified candidate member nodes and candidate organization partitions Tags and Industry Portfolio A set of objects that simultaneously meet resource tags and industry conditions Simplified candidate member nodes and candidate organization partitions In this embodiment, a candidate resource node refers to a resource tag node or resource item node that matches the target resource conditions, a candidate member node refers to a user node or business card node that is directly associated with the target resource conditions, and a candidate organization partition refers to an organization member partition that contains matching members or matching resource tags in a highly connected organization node.

[0048] When determining candidate resource nodes, the system reads resource tag nodes in the heterogeneous network resource map according to standard resource tags or resource categories, and filters out resource nodes that are invalid, expired, or obviously mismatched with regional or industry constraints, thereby forming a set of candidate resource nodes that can be used as the basis for target resource parties or path endpoints.

[0049] When determining candidate personnel nodes, the system reads user nodes or business card nodes that are directly associated with the target resource conditions based on the resource inverted index, and filters them by combining personnel status, business card status, resource tag validity status and relationship visibility status, so that candidate personnel nodes have at least a resource supply relationship or a provable resource association record corresponding to the target resource conditions.

[0050] When determining candidate organization partitions, the system matches the compressed index of organization member partitions established in S2 based on the target resource conditions, and prioritizes retaining organization member partitions whose resource tag set, region dimension, and industry dimension are consistent with the target resource conditions. This allows subsequent path searches to directly reference candidate organization partitions to generate virtual bridge edges when they find highly connected organization nodes.

[0051] The table below shows the candidate object generation rules, which are used to explain the source and retention conditions of the candidate set after determining the candidate resource nodes, candidate member nodes, and candidate organization partitions through the resource inverted index.

[0052] Candidates Data source Reservation conditions Subsequent effects Candidate resource nodes Resource tag node or resource item node Resource tags or resource categories match the target resource conditions As a basis for judging target resource parties Candidate node User node or business card node Directly associated with target resource conditions and in a valid state As a reachable target node, it participates in path search. Candidate organization partitions Organization member partition compressed index Matching zone resource tags, regions, or industries with target resource conditions Virtual bridging processing for highly connected organizational nodes Excluded objects Expired resources, invalid members, expired relationships, or invisible relationships Validity or constraint conditions are not met. Do not proceed to subsequent path search After generating the candidate set, the system performs deduplication and association verification on the candidate resource nodes, candidate member nodes, and candidate organization partitions. When the same user has multiple business cards, the business card node with a higher matching degree with the target resource conditions can be retained. When the same organization has multiple similar partitions, the organization member partitions that simultaneously meet the resource tag, region, and industry constraints can be retained.

[0053] Through the above processing, the execution result of S3 is to obtain the candidate resource node set, candidate member node set, and candidate organization partition set corresponding to the path retrieval request. The candidate set is used to limit the target scope of subsequent path search, so that the system does not blindly expand outward from the starting user, but performs relational path retrieval under the constraints of target resource conditions.

[0054] S4. In this embodiment, after obtaining the set of candidate resource nodes, the set of candidate personnel nodes, and the set of candidate organization partitions, the system performs a path search with a limited number of hops starting from the starting user. Specifically, the user node corresponding to the starting user is used as the initial node, and the initial node, the set of visited nodes, the set of visited relation edges, the current number of hops, the current path score, the number of organizational bridges already passed, the remaining expansion budget, and the path signature are written into the path search queue.

[0055] The following table is a path search status table, which describes the data content saved by the system for each path to be expanded when the user performs a path search with a limited number of hops from the starting point.

[0056] Field Name Technical meaning Processing purpose Current node Currently, user nodes, business card nodes, organization nodes, or resource tag nodes that need to be expanded Determine the next round of expansion objects Visited node set The set of nodes already visited in the current path Avoid repeatedly accessing the same node The set of relation edges has been passed. The set of relation edges already traversed in the current path Form an interpretable path link Current hop count The number of edges traversed from the starting user to the current node. Determine if the limit number of jumps has been exceeded. Current path score Provisional evaluation results of the current path in terms of resource matching and relationship strength For subsequent sorting and pruning calls Number of organizational bridging sessions already completed The number of times the current path passes through the highly connected organization nodes Control the number of tissue bridges Remaining expansion budget The current search still allows for expansion of the number of organizations, partitions, or members. Controlling query computation volume Path signature Path markers formed by node sequences, edge type sequences, and organizational partition identifiers Used for candidate path deduplication When the system performs a path search with a limited number of hops starting from the user, it reads the current path status from the path search queue. It first determines whether the current number of hops has reached the maximum number of hops in the path retrieval request. If the current number of hops has reached the maximum number of hops and the current node has not yet hit a candidate resource node or candidate member node, it stops further expanding the path.

[0057] When the current node is a user node, business card node, or resource tag node, the system expands the ordinary node according to the valid relationship edges. Specifically, it reads the business card exchange edge, organization affiliation edge, resource supply edge, cooperation proof edge, and recommendation relationship edge connected to the current node, and filters the relationship edges whose edge status is deleted, invalid, expired, or invisible.

[0058] When expanding ordinary nodes according to effective relation edges, the system also judges whether adjacent nodes have the value of further expansion based on target resource conditions, regional constraints, industry constraints, and edge type sequence constraints. If there is no direct or indirect relationship between the adjacent node and the candidate resource node set, candidate member node set, or candidate organization partition set, the priority of the adjacent node entering the path search queue is reduced or it is no longer expanded.

[0059] The table below shows the rules for expanding ordinary nodes, which is used to explain how the system processes different types of valid relation edges.

[0060] Relationship edge type Extended Source Extended Object Processing rules Business card exchange side User node or business card node User nodes or business card nodes that have exchanged business cards Expansion is allowed when the state is valid and visible. Organizational affiliation border User node or business card node Belonging organization node Used to determine whether to enter a regular organization node or a highly connected organization node. Resource supply side User node or business card node Resource tag nodes Prioritize retention when matching target resource conditions Cooperation Proof of Edge User node, business card node, or collaboration event node There are associated nodes with proof of cooperation. As a proof of relation, it participates in subsequent sorting. Recommendation of relationship edge User node or business card node User nodes or business card nodes that can be introduced Participating in the search as a transferable path When the current node is an organization node, the system first reads the high connection marker of the organization node. If the organization node is not marked as a high connection organization node, the system still expands the ordinary node according to the valid relation edges, that is, it reads the members under the organization node that can participate in the retrieval according to the valid organization membership edges, but is limited by the maximum relation hop count and the remaining expansion budget.

[0061] When the current node is a high-connectivity organization node, the system generates a virtual bridge edge for the high-connectivity organization node based on the matched candidate organization partitions. Specifically, the organization identifier of the high-connectivity organization node is matched with the set of candidate organization partitions obtained in S3. Only when there is a candidate organization partition belonging to the high-connectivity organization node and matching the target resource conditions, a logical connection is generated to indicate that the organization node continues to search for candidate members through the corresponding organization member partition.

[0062] Virtual bridging edges are not real relationship edges pre-stored in the heterogeneous network resource graph. Instead, they are path state records temporarily generated when virtual bridging edges are generated for highly connected organizational nodes based on the hit candidate organizational partitions. They are used to advance the current path from the highly connected organizational node to the organizational member partition level related to the target resource conditions without immediately reading all members.

[0063] The table below is a virtual bridge edge status table, which describes the data recorded by the system when generating virtual bridge edges in S4.

[0064] Field Name Technical meaning Subsequent uses Organizational Identification The unique identifier of the current high-connectivity organization node Determine the organization to which the virtual bridging edge belongs. Partition identifier Unique identifier of the candidate organization partition Determine the scope of members who can be deployed in a controlled manner. Number of candidate members Number of members that can participate in the search within the member partition of the target organization Determine the cost of subsequent deployment Partition matching degree The degree of matching between the target organization's member partitions and the target resource conditions. Determine whether to enter controlled deployment Organizational bridging costs The cost of continuing the search through this highly connected organization node and organization member partitions Participate in subsequent path sorting and dominant pruning Partition version number The current update of the hit organization member section Used for cache validation and path validity determination In S4, the system does not fully expand the members of the highly connected organization node. Specifically, it does not read all the member nodes corresponding to the organizational membership edges under the highly connected organization node, nor does it connect the members within the organization pairwise to generate candidate paths. Instead, it only generates virtual bridging edges corresponding to the partitions of the hit candidate organizations, and uses these virtual bridging edges as input for the subsequent S5 to determine the resource matching conditions and expansion budget conditions.

[0065] For example, when the path search finds a chamber of commerce node with a large number of members, if the target resource condition is corporate financing resources, the system will not traverse all members under the chamber of commerce. Instead, it will first search for candidate organization partitions in the chamber of commerce that match the financing resource label, target region, and target industry, and then generate virtual bridge edges for highly connected organization nodes based on the matched candidate organization partitions.

[0066] Through the above processing, S4 obtains an intermediate path set that includes the expansion results of ordinary nodes and the state of virtual bridge edges. The intermediate path set not only preserves the reachability from the starting user to the target resource, but also limits the path explosion caused by the high-connectivity organization node by not fully expanding the members of the high-connectivity organization node.

[0067] S5. In this embodiment, the system takes the virtual bridge edge status generated in S4 as input, performs condition verification on each virtual bridge edge, and only allows reading specific member nodes from the organization member partition corresponding to the virtual bridge edge when it is confirmed that the virtual bridge edge meets the resource matching conditions and expansion budget conditions.

[0068] Specifically, the system determines whether to proceed with the process of selecting candidate members from the corresponding organization's member partitions as the next hop node to continue the search, based on the target resource conditions, the partition matching degree recorded by the virtual bridge edge, the current path hop count, the remaining expansion budget, the number of candidate members, and the member validity status.

[0069] The table below shows the virtual bridging edge expansion condition table, which explains the specific conditions used by the system to determine whether a virtual bridging edge meets the resource matching condition and expansion budget condition, and the handling method when the condition is not met.

[0070] Condition Category Judge content Process when satisfied Handling when not satisfied Resource matching conditions Matching organizational members' resource tags, regions, or industries with target resource conditions. Retain the virtual bridging edge Delete the virtual bridge edge or reduce its expansion priority. Partition Validity Conditions The organization member partition status is valid and the partition version number is available. Allow entry to candidate member reading Do not read members within this partition Jump count condition The current path hop count has not exceeded the hop limit. Allow continued expansion to member nodes Stop the path from expanding further. Organization to develop budget The number of highly connected tissues allowed for this search has not yet been exhausted. Preserve organizational bridging status Stop continuing through this organization Split budget The number of organization member partitions that can be expanded during this search has not yet been exhausted. Preserve hit partition No longer read new organization member partitions Members develop budget The number of candidate members allowed to be read in this search has not yet been exhausted. Read a controlled number of candidate members Stop reading more members Candidate member availability There are members in the partition that are in a valid state and associated with the target resource conditions. Forming candidate nodes for the next hop Delete the virtual bridge edge Regarding resource matching conditions, the system does not require all members in the organization member partition to be consistent with the target resource conditions. Instead, it determines whether the resource tag set, region dimension, or industry dimension of the organization member partition has a matching relationship with the target resource conditions in the path retrieval request, thereby determining whether the resource matching part of the virtual bridging edge satisfies the resource matching conditions and the expansion budget conditions.

[0071] For the expansion budget condition, the system records the remaining number of organization expansions, partition expansions, and member expansions in this path retrieval request. Only when all three types of budgets are not exhausted will the system select candidate members from the corresponding organization member partitions as the next hop node to continue the search, so as to avoid the continuous spread of highly connected organization nodes in a single retrieval.

[0072] During the candidate member reading process, the system does not traverse all members in the corresponding organization member partition. Instead, it first reads the candidate member list in the organization member partition and filters them according to member status, resource association, relationship visibility status, whether they have been accessed by the current path, and candidate member sorting results, so as to select candidate members from the corresponding organization member partition as the next hop node to continue the search.

[0073] The table below shows the candidate member selection rules, which explains how the system selects candidate members from the organization's member partitions who can serve as the next-hop node.

[0074] order Processing objects Filtering rules Output 1 Partition Member Index List Read member nodes and business card nodes that are valid or eligible for retrieval. Preliminary selection members 2 Preliminary selection members Exclude nodes that have already been visited in the current path Extensible member collection 3 Extensible member collection Members who are directly or indirectly associated with the target resource conditions are retained. Resource matching member set 4 Resource matching member set The reading order is determined by the pre-sorting of the candidate member list. Sort member set 5 Sort member set Extract members that do not exceed the maximum allowed number of expanded members. Next hop candidate node set When the same highly connected organizational node hits multiple candidate organizational partitions, the system determines whether the virtual bridge edge corresponding to each candidate organizational partition meets the expansion conditions, and prioritizes processing virtual bridge edges with higher resource matching degree, lower organizational bridging cost and a controllable number of candidate members, so as to ensure that the actual member expansion is generated only when the virtual bridge edge meets the resource matching conditions and expansion budget conditions.

[0075] When the same candidate member exists in multiple organizational member partitions, the system retains the partition source that can form a shorter path, has a stronger relationship proof, or has a lower organizational bridging cost, and records the duplicate source as path candidate information without adding it to the path search queue again. This ensures that the result of selecting a candidate member as the next hop node from the corresponding organizational member partition to continue the search will not produce duplicate paths due to partition intersection.

[0076] The system will write the selected candidate member nodes, their corresponding virtual bridge edges, the set of relation edges that have been traversed on the current path, the updated current hop count, the remaining expansion budget, and the path signature into the new path state. This will enable the candidate members to participate in subsequent searches as the next hop nodes of the current path, thereby completing the selection of candidate members from the corresponding organization member partitions as the next hop nodes to continue the search.

[0077] Through the above processing, the execution result of S5 is to convert virtual bridging edges that meet the resource matching conditions and expansion budget conditions into a controlled number of next-hop candidate nodes, and to block virtual bridging edges that do not meet the conditions at the organizational member partition level, so that the system can limit the full expansion of members caused by highly connected organizational nodes while maintaining the accessibility of network resource paths.

[0078] S6. In this embodiment, after the system forms the expansion results of ordinary nodes and the next-hop candidate nodes after controlled expansion, it prunes and sorts the candidate paths. Pruning is used to delete candidate paths that do not meet the retrieval constraints or have low actual transfer value, and sorting is used to determine the priority output order among the remaining candidate paths.

[0079] The system first performs basic pruning on each candidate path, which includes determining whether the candidate path exceeds the limit number of hops, whether the candidate path contains relation edges with a status of deleted, invalid, expired or invisible, whether the candidate path repeatedly passes through the same node, and whether the edge type sequence in the candidate path conforms to the preset relation path rules.

[0080] The table below shows the basic pruning rules for candidate paths, which describes the pruning rules that the system first performs when pruning and sorting candidate paths.

[0081] Pruning types Determine the object Judgment rules Processing results Maximum hop count pruning Current path hop count The current path has exceeded the hop limit. Stop further expansion and delete the candidate path. Edge state pruning relation edge state There are deleted, invalid, expired, or invisible relationship edges. Delete the candidate path Repeated pruning of nodes Visited node set The same candidate path repeatedly passes through the same node Delete duplicate extended parts or delete the candidate path. Edge type sequence pruning Sequence of relation edge types in the path Does not conform to the pre-defined interpretable relation sequence Delete the candidate path Resource matching and pruning Target resource conditions and path endpoints or intermediate members No target resource association exists Delete or reduce the priority of the candidate path Organizational budget pruning Number of organizations, divisions, and members Exceeding the expansion budget for this search Stop expanding the candidate path When performing edge type sequence pruning, the system allows user nodes to form interpretable paths via business card exchange edges, organization affiliation edges, virtual bridging edges, resource supply edges, or cooperation proof edges, and prohibits path structures that do not have the meaning of transition interpretation, such as resource label nodes continuously jumping to unrelated resource label nodes or organization nodes jumping to member nodes unrelated to the target resource.

[0082] After basic pruning, the system continues to perform path domination pruning. Path domination pruning means that for multiple candidate paths to the same target resource or the same candidate node, if one candidate path is no higher than another candidate path in terms of path hop count, organizational bridging cost, and invalidity risk, and is no lower than another candidate path in terms of resource matching degree, relationship strength, and relationship proof status, then the latter candidate path will be deleted or its output priority will be reduced.

[0083] The following table is a path dominance pruning judgment table, which is used to explain the main dimensions used by the system when comparing two candidate paths.

[0084] Comparison Dimensions Priority retention direction Technical meaning Path hop count Fewer jumps This indicates a shorter connection layer from the originating user to the target resource provider. Organizational bridging costs Lower cost This indicates that weak connections generated through highly connected organizational nodes are less costly. Resource matching degree Higher matching degree This indicates that the endpoint or intermediate candidate member of the path is more suitable for the target resource conditions. Relationship strength Higher strength This indicates a more stable relationship through business card exchanges, collaborations, or referrals. Relationship Proof Status More sufficient proof This indicates evidence of business card exchange, collaborative events, or being present at the same event. Risk of ineffectiveness Lower risk This indicates a lower risk of the path containing invalid, expired, or version-changed data. In this embodiment, the system generates a path signature for each candidate path. The path signature is determined by the node sequence, the relation edge type sequence, and the identifier of the organization member partitions traversed. The system uses the path signature to identify duplicate paths and similar paths, thereby avoiding the generation of a large number of candidate paths with different forms but the same actual transition meaning by multiple similar partitions under the same highly connected organization node.

[0085] After pruning, the system continues to sort the candidate paths and determines the ranking results based on resource matching degree, relationship strength, relationship proof status, path hop count, and organizational bridging cost. The higher the resource matching degree, the higher the relationship strength, and the more sufficient the relationship proof status, the higher the priority of the candidate path. The higher the path hop count and the higher the organizational bridging cost, the lower the priority of the candidate path.

[0086] To facilitate the explanation of the sorting criteria, this embodiment sets up the following candidate path sorting factor table, which is used to explain the technical meaning of the sorting part in the pruning and sorting of candidate paths.

[0087] sorting factor Technical meaning Impact on sorting Resource matching degree The degree of matching between the path endpoint or candidate member and the target resource conditions. The higher the match, the higher the ranking. Relationship strength The stability of relationships such as business card exchange, recommendation, and cooperation within the path. The higher the strength, the higher the ranking. Relationship Proof Status Are there any records of business card exchanges, collaborative events, or events held together? The more sufficient the proof, the higher the ranking. Path hop count Number of connections from the originating user to the target resource provider The fewer the number of jumps, the higher the ranking. Organizational bridging costs Bridging costs arising from high-connectivity organization nodes and organization member partitions The lower the cost, the higher the ranking. Data freshness Update time of relationship edges, resource tags, or organization member sections The newer the data, the higher it ranks in the sorting. In one optional implementation, the system may use a rule-based scoring method to rank candidate paths. The scoring is used only to illustrate the ranking logic in this embodiment, specifically as follows: PathScore = a × ResourceMatch + b × RelationStrength + c × ProofLevel + d × Freshness - e × HopCount - f × BridgeCost, where PathScore represents the candidate path score, ResourceMatch represents the resource matching degree, RelationStrength represents the relationship strength, ProofLevel represents the relationship proof status, Freshness represents the data freshness, HopCount represents the path hop count, BridgeCost represents the organizational bridging cost, and a, b, c, d, e, and f represent the configurable weights of the corresponding ranking factors.

[0088] When the system does not use the above scoring formula, it can also use a hierarchical sorting rule. For example, it can output direct business card relationship paths first, then output organizational bridging paths with proof of cooperation, and then output ordinary organizational member relationship paths, and sort them according to the number of path hops and organizational bridging costs within the same level.

[0089] After the system completes pruning and sorting, it retains the candidate paths that are ranked first and have not been pruned, and writes the node sequence, relation edge type, organization information passed through, target resource label, path hop count, relation proof summary, and whether virtual bridging expansion is used for each candidate path into the set of paths to be output.

[0090] Through the above processing, S6 obtains a set of output paths after pruning and sorting. The set of output paths reduces duplicate paths, invalid paths, and weak connection noise paths, and prioritizes retaining personal network resource relationship paths with high resource matching degree, sufficient relationship proof, and low organizational bridging cost.

[0091] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A relationship path retrieval method for a network resource map, characterized in that: include: S1. Acquire user, business card, organization, resource tags and cooperation event data to construct a heterogeneous network resource map; S2. Identify high-connectivity organizational nodes based on the number of organizational nodes' members, the number of connected edges, or historical expansion overhead, and build an organizational member partitioned compressed index; S3. Receive a path retrieval request containing the starting user and target resource conditions, and determine candidate resource nodes, candidate user nodes and candidate organization partitions through the resource inverted index; S4. Perform a path search with a limited number of hops starting from the user. For ordinary nodes, expand the edge based on the effective relationship. For high-connection organization nodes, generate virtual bridge edges based on the hit candidate organization partitions, without fully expanding the members of the high-connection organization node. S5. When the virtual bridge edge meets the resource matching condition and the expansion budget condition, select a candidate member from the corresponding organization member partition as the next hop node to continue the search; S6. Prune and sort the candidate paths, and output the network resource relationship path from the starting user to the target resource party.

2. The relationship path search method for the contact resource map according to claim 1, characterized by: The nodes in the heterogeneous network resource graph include user nodes, business card nodes, organization nodes, resource tag nodes, and cooperation event nodes. The relationship edges in the heterogeneous network resource graph include at least two of the following: business card exchange edges, organization affiliation edges, resource supply edges, cooperation proof edges, and recommendation relationship edges. Each relationship edge is configured with at least one of the following attributes: edge type, edge status, relationship strength, proof level, and update version.

3. The relationship path search method for the contact resource map according to claim 1, characterized by: When identifying the highly connected organizational node, the number of members, the number of connection edges, and the number of times or time spent expanding members in historical queries are counted. When any statistical result reaches the corresponding preset threshold, the organizational node is marked as a highly connected organizational node and is used as an organizational node that needs to undergo organizational member partitioning and compression indexing.

4. The relationship path retrieval method for a network resource map according to claim 1, characterized in that: The organizational member partition compression index partitions members under the same highly connected organizational node according to at least two of the following dimensions: region, industry, role, resource tag, member status, active time, and relationship proof. Each organizational member partition records a partition identifier, the organization identifier, the number of members in the partition, a partition member index list, a candidate member list, a resource tag set, and a partition version number.

5. The relationship path retrieval method for a network resource map according to claim 1, characterized in that: Determining candidate resource nodes, candidate personnel nodes, and candidate organizational partitions through a resource inverted index includes: converting the target resource conditions into standard resource tags or resource categories, and using the standard resource tags or resource categories in combination with at least one of regional constraints and industry constraints as search conditions to query resource tag nodes that match the target resource conditions, personnel nodes that are directly associated with the target resource conditions, and organizational member partitions containing matching members.

6. The relationship path retrieval method for a network resource map according to claim 1, characterized in that: The virtual bridging edge is a logical edge temporarily generated during the path search process based on the highly connected organization node and its matched candidate organization partitions. The virtual bridging edge is used to represent the path status of continuing to search for candidate members from the highly connected organization node through the corresponding organization member partition. The virtual bridging edge records the organization identifier, partition identifier, number of candidate members, partition matching degree, organization bridging cost, and the upper limit of the number of members allowed to be expanded.

7. The relationship path retrieval method for a network resource map according to claim 1, characterized in that: The virtual bridge edge satisfies the resource matching conditions and expansion budget conditions as follows: the matching degree between the corresponding organization member partition and the target resource condition reaches a preset matching threshold, the current path hop count does not exceed the limit hop count, the organization expansion budget, partition expansion budget and member expansion budget corresponding to the current path retrieval request are not exhausted, and there are candidate members in the corresponding organization member partition whose status is valid and associated with the target resource condition.

8. The relationship path retrieval method for a network resource map according to claim 1, characterized in that: Pruning candidate paths includes: pruning candidate paths that exceed the maximum hop count limit; pruning candidate paths containing invalid relation edges; pruning candidate paths that repeatedly pass through the same node; pruning candidate paths that do not conform to the preset edge type sequence; and pruning candidate paths that reach the same target resource and are dominated by other candidate paths in terms of path hop count, organizational bridging cost, resource matching degree, and relation strength.

9. The relationship path retrieval method for a network resource map according to claim 1, characterized in that: The ranking of candidate paths includes: determining the ranking result of candidate paths based on resource matching degree, relationship strength, relationship proof status, path hop count and organization bridging cost; when outputting the network resource relationship path, at least the path node sequence, relationship edge type, traversed organization information, target resource tag, path hop count and relationship proof summary are output.

10. A relationship path retrieval system for a network resource map, characterized in that: The method for retrieving relationship paths according to any one of claims 1-9 includes a processor, a memory, and a data access unit, a heterogeneous graph construction unit, a high-connectivity organization identification unit, an organization compression index unit, a resource candidate retrieval unit, a path search unit, a virtual bridging processing unit, a pruning and sorting unit, and a path output unit, all stored in the memory and invoked by the processor. The heterogeneous graph construction unit is used to construct a heterogeneous network resource graph containing users, business cards, organizations, resource tags, and collaborative events. The high-connectivity organization identification unit and the organization compression index unit are used to identify high-connectivity organization nodes and establish a compressed index for organization member partitions. The resource candidate retrieval unit is used to obtain candidate resource nodes, candidate member nodes, and candidate organization partitions based on target resource conditions. The path search unit and the virtual bridging processing unit are used to generate virtual bridging edges and controllably expand candidate members when a high-connectivity organization node is found in the path search. The pruning and sorting unit and the path output unit are used to prune and sort the candidate paths and output the network resource relationship path.

Images