Intelligent matching method and system for scientific research talents and R&D tasks based on knowledge graph

Abstract

The present invention provides a method and system for intelligently matching scientific research talents and R&D tasks based on a knowledge graph. By acquiring and parsing a plurality of candidate R&D tasks and their matching priorities, and combining the research-related knowledge feature data between scientific research talent nodes, a plurality of scientific research talent knowledge sub-graphs are constructed. Furthermore, the scientific research talent nodes are prioritized according to the task matching priority, so that in the same scientific research talent knowledge sub-graph, scientific research talent nodes of different sequences have different matching priority information. In the matching process, according to the matching priority information of the target talent knowledge entity, the corresponding target R&D task is quickly determined, and the target R&D task is associated with the corresponding talent knowledge entity, thereby not only improving the matching efficiency of scientific research talents and R&D tasks, but also ensuring the accuracy and rationality of the matching, which is helpful to optimize the allocation of scientific research resources and promote the smooth progress of scientific research projects.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and system for intelligently matching scientific research talents and R&D tasks based on a knowledge graph. Background Art

[0002] In the field of scientific research, the key to ensuring the success of scientific research projects is to reasonably and efficiently match scientific research talents with R&D tasks. Traditional methods of matching scientific research talents with tasks mostly rely on manual experience judgment, or are based only on simple keyword matching methods. These methods have problems such as low matching efficiency and inaccurate matching results. Especially when facing large-scale, multi-field scientific research projects, traditional methods are even more inadequate.

[0003] Among related technologies, with the development of knowledge graph technology, its application in the field of scientific research has gradually attracted attention. Knowledge graphs can structuredly represent entities and their relationships, providing new ideas for matching scientific research talents with R&D tasks. However, most existing knowledge graph-based matching methods only focus on the attribute characteristics of scientific research talents and R&D tasks themselves, ignoring the complex research associations between scientific research talents, resulting in certain limitations in matching results. Summary of the invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a method for intelligently matching scientific research talents and R&D tasks based on a knowledge graph, the method comprising:

[0005] Obtain matching priorities of multiple candidate R&D tasks and multiple previously defined tasks;

[0006] Acquire multiple scientific research talent nodes and research-related knowledge feature data between the scientific research talent nodes for matching R&D tasks, and determine multiple scientific research talent knowledge sub-graphs based on the research-related knowledge feature data;

[0007] Determine the matching priority information of each scientific research talent node in the multiple scientific research talent knowledge sub-graphs according to the multiple task matching priorities, wherein the matching priority information of scientific research talent nodes in different sequences in the same scientific research talent knowledge sub-graph is different;

[0008] When determining a matching progress that is suitable for matching a research and development task for a target talent knowledge entity, determining a target research and development task corresponding to the target talent knowledge entity according to matching priority information of the target talent knowledge entity;

[0009] The target R&D task is associated with the target talent knowledge entity.

[0010] On the other hand, an embodiment of the present invention also provides an intelligent matching system for scientific research talents and R&D tasks based on knowledge graphs, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0011] Based on the above aspects, the embodiments of the present application can efficiently achieve accurate matching of scientific research talents and R&D tasks. By acquiring and parsing multiple candidate R&D tasks and their matching priorities, combined with the research-related knowledge feature data between scientific research talent nodes, multiple scientific research talent knowledge sub-graphs are constructed. Furthermore, the scientific research talent nodes are prioritized according to the task matching priority, so that in the same scientific research talent knowledge sub-graph, scientific research talent nodes of different sequences have different matching priority information. In the matching process, according to the matching priority information of the target talent knowledge entity, the corresponding target R&D task is quickly determined, and the target R&D task is associated with the corresponding talent knowledge entity, thereby not only improving the matching efficiency of scientific research talents and R&D tasks, but also ensuring the accuracy and rationality of the matching, which helps to optimize the allocation of scientific research resources and promote the smooth progress of scientific research projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a schematic diagram of the execution flow of the method for intelligently matching scientific research talents and R&D tasks based on knowledge graph provided in an embodiment of the present invention.

[0013] Figure 2 It is a schematic diagram of the hardware architecture of the intelligent matching system for scientific research talents and R&D tasks based on knowledge graph provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 It is a flow chart of a method for intelligently matching scientific research talents and R&D tasks based on knowledge graph provided by an embodiment of the present invention. The method for intelligently matching scientific research talents and R&D tasks based on knowledge graph is introduced in detail below.

[0015] Step S110, obtaining a plurality of candidate R&D tasks and a plurality of previously defined task matching priorities.

[0016] In this embodiment, the server is the central task allocation and management server of the scientific research platform. In this scientific research platform, there are multiple different R&D projects, which generate numerous candidate R&D tasks. For example, the scientific research platform is carrying out a new drug R&D project, and the candidate R&D tasks include "new drug molecular structure design", "drug synthesis process optimization", "drug clinical trial plan formulation", etc. At the same time, the scientific research platform also has an artificial intelligence algorithm improvement project, and its candidate R&D tasks include "algorithm model architecture optimization", "algorithm training data expansion and cleaning", "algorithm performance evaluation indicator system construction", etc.

[0017] The task matching priority is defined in advance based on the strategic goals of the scientific research platform, the urgency of the project, and the allocation of resources. Taking the drug development project as an example, if the scientific research platform wants to launch a new anti-cancer drug as soon as possible to seize the market, then the task of "new drug molecular structure design" may be given a higher task matching priority because it is a key preliminary step in drug development. For the artificial intelligence algorithm improvement project, if the scientific research platform has just obtained a batch of new computing resources and urgently needs to use these resources to optimize the existing algorithm model to improve computing efficiency, then the task matching priority of "algorithm model architecture optimization" will be higher. The server obtains these candidate R&D tasks and their corresponding task matching priority information by accessing the project management database within the scientific research platform. This database stores the detailed planning, task lists, and task priority settings of each project jointly determined by the senior management of the scientific research platform and the project leader.

[0018] Step S120, obtaining multiple scientific research talent nodes and research-related knowledge feature data between the scientific research talent nodes for matching R&D tasks, and determining multiple scientific research talent knowledge sub-graphs based on the research-related knowledge feature data.

[0019] In this embodiment, in this scientific research platform, the server obtains relevant information of scientific research talents from multiple data sources to determine scientific research talent nodes. For example, the human resources management system within the scientific research platform records the basic information of each scientific researcher, including academic qualifications, majors, work experience, published papers, etc. At the same time, the server also obtains information such as the participation of scientific researchers in various projects and the roles they assume from the scientific research project management system of the scientific research platform. These scientific researchers constitute multiple scientific research talent nodes.

[0020] Taking the drug R&D team and the AI ​​algorithm R&D team as examples, the researchers in the drug R&D team include medicinal chemists, pharmacologists, clinical researchers, etc., and the researchers in the AI ​​algorithm R&D team include algorithm engineers, data scientists, mathematical modeling experts, etc. The server regards these personnel as scientific research talent nodes.

[0021] To determine the research-related knowledge feature data between scientific research talent nodes, the server first obtains multi-source academic research data. For example, the server is connected to the academic paper database purchased by the scientific research platform, which contains a large number of academic papers in the fields of drug development and artificial intelligence worldwide; the scientific research project database records detailed information on internal and external cooperation projects of the scientific research platform; the patent database provides existing patent technology information in related fields; the scholar social network data can reflect the academic exchanges and cooperation intentions between scientific researchers; and the institutional cooperation data shows the cooperation between the scientific research platform and other scientific research institutions on various projects.

[0022] These multi-source academic research data are processed using natural language processing algorithms and entity recognition algorithms. Taking an academic paper on the development of new anti-cancer drugs as an example, the natural language processing algorithm will analyze the content of the paper, and the entity recognition algorithm will identify key information such as the names of researchers, research institutions, and research results in the paper. By processing a large amount of such data, a list of scientific research talent nodes is generated. Each scientific research talent node contains a unique identifier (such as employee number) and basic information (such as name, professional field, etc.).

[0023] For each scientific research talent node, the server extracts the scientific research content features of the scientific research results data of the node through text mining and semantic analysis algorithms. For example, for a medicinal chemist, the keywords that frequently appear in his published papers (such as "drug molecular structure" and "drug active ingredients") form a keyword feature vector; the paper topic (such as "molecular design strategy of anticancer drugs") forms a topic feature vector; the research field (such as "medicinal chemistry") forms a research field feature vector.

[0024] Construct a cooperation network diagram between scientific research talents based on the cooperation information in multi-source academic research data. For example, if a medicinal chemist and a pharmacologist have a cooperative relationship in a drug development project, there will be a corresponding connection link in the cooperation network diagram. Combine the scientific research content characteristics and the cooperation network diagram to calculate the research association knowledge characteristics between scientific research talent nodes. For example, the research association strength between two researchers who have collaborated in multiple drug development projects and have highly related research fields (similar scientific research content characteristics) will be higher, and generate a feature data matrix containing the research association strength between scientific research talent nodes.

[0025] The server uses linear discriminant analysis and recursive feature elimination algorithms to optimize the feature data matrix. For example, there may be some redundant features in the feature data matrix or features that do not contribute much to distinguishing the relationship between scientific research talents. These algorithms can remove these unnecessary features and generate an optimized feature data matrix. Then, the optimized feature data matrix is ​​verified using actual scientific research cooperation data as a verification set. For example, based on the actual scientific research cooperation within the scientific research platform, if the optimized feature data matrix shows that two researchers have a high degree of correlation, but in fact they have rarely cooperated in the past, then it is necessary to further adjust the matrix based on the verification results, and after multiple adjustments, generate research correlation knowledge feature data between scientific research talent nodes.

[0026] The basic scientific research talent knowledge graph is constructed based on the research-related knowledge feature data between scientific research talent nodes. In this graph, each scientific research talent node is a knowledge entity, and there is research-related knowledge feature data between the two scientific research talent nodes associated with the entity association link. For example, in the basic scientific research talent knowledge graph in the field of drug research and development, the link between medicinal chemists and pharmacologists indicates that there is an association between them based on research-related knowledge feature data (such as joint participation in projects, complementary research fields, etc.). Then, the basic scientific research talent knowledge graph is divided according to the previously defined knowledge graph division strategy to generate multiple scientific research talent knowledge sub-graphs. For example, it can be divided according to research fields (such as drug research and development sub-graphs, artificial intelligence algorithm research and development sub-graphs) or according to project teams (such as the sub-graph corresponding to a certain anti-cancer drug research and development project team, and the sub-graph corresponding to a certain artificial intelligence algorithm optimization project team).

[0027] Step S130, determining the matching priority information of each scientific research talent node in the multiple scientific research talent knowledge sub-graphs according to the multiple task matching priorities, and the matching priority information of the scientific research talent nodes in different sequences in the same scientific research talent knowledge sub-graph is different.

[0028] Take the previously divided drug development sub-graph and artificial intelligence algorithm development sub-graph as examples. In the drug development sub-graph, assume that the number of knowledge entities (scientific research talent nodes) in this sub-graph is 10, including medicinal chemists, pharmacologists, clinical researchers and other scientific researchers of different professions. The server divides these knowledge entities into the first knowledge entity sequence and the second knowledge entity sequence. Since the number is divisible by 2, it is evenly distributed. For example, the first knowledge entity sequence includes medicinal chemists, clinical researchers, etc., and the second knowledge entity sequence includes other personnel such as pharmacologists.

[0029] Construct a unique mapping relationship between the first knowledge entity sequence and the second knowledge entity sequence. For example, a medicinal chemist (an entity in the first knowledge entity sequence) and a pharmacologist (an entity in the second knowledge entity sequence) have a specific mapping relationship, which may be based on their order or mutual collaboration in the drug development process.

[0030] The server sorts and integrates multiple task matching priorities to generate multiple matching priority integration sequences. For example, for the task matching priorities in a drug development project, after sorting and integration, a matching priority integration sequence such as "new drug molecular structure design" > "drug synthesis process optimization" > "drug clinical trial plan formulation" may be obtained.

[0031] The matching priority integration sequence corresponding to each first knowledge entity is determined from these matching priority integration sequences. For example, for the first knowledge entity of medicinal chemist, since it plays a key role in drug molecular structure design, the corresponding matching priority integration sequence may be that the task of "new drug molecular structure design" has the highest priority.

[0032] The matching priority information corresponding to each first knowledge entity is determined according to the matching priority integration sequence corresponding to each first knowledge entity. For example, for a medicinal chemist, the high priority of the task "new drug molecular structure design" is used as its matching priority information, indicating that it has a higher matching degree when performing research and development work related to this task.

[0033] The matching priority information corresponding to each first knowledge entity is reversely converted to generate the matching priority information of the second knowledge entity corresponding to each first knowledge entity. For example, for the pharmacologist (second knowledge entity) corresponding to the medicinal chemist, the tasks are arranged in ascending order of matching priority according to the matching priority information of the medicinal chemist, and the reverse order obtained is used as the matching priority information of the pharmacologist. That is, if the medicinal chemist has a high priority in the task of "new drug molecular structure design", then for the pharmacologist, the priority of the subsequent tasks related to this task (such as drug activity testing, etc.) may be adjusted accordingly.

[0034] In another case, take the labeling process to determine the matching priority information as an example. The server labels the drug development sub-map and generates a labeled scientific research talent knowledge sub-map. Obtain a set of labeling basic information from data sources such as the subject classification system library, the terminology standard library in the scientific research field, and the industry general research direction guide library. For the research and development data of the knowledge entities (scientific researchers) in the drug development sub-map, preliminary labeling is performed based on the labeling basic information. For example, for the research and development work of medicinal chemists, if it mainly involves work in organic chemical synthesis, preliminary labeling is performed based on the organic chemical classification information in the subject classification system library.

[0035] Based on the preliminary annotated scientific research talent knowledge sub-graph set and the annotated basic information set, the annotation basis of the entity association relationship is determined to generate a relationship annotation basis set. For example, if there is a sequential relationship between medicinal chemists and pharmacologists in the R&D process (medicinal chemists synthesize drug molecules first, and pharmacologists then test drug activity), this relationship will be built into the rule system in the relationship annotation basis set.

[0036] According to the relationship annotation, the entity association relationship in each scientific research talent knowledge sub-graph is annotated according to the rules in the set. For example, the association relationship between medicinal chemists and pharmacologists is annotated as a "synthesis-testing" relationship, indicating their sequential relationship in the drug development process.

[0037] Assume that the number of knowledge points M annotated by knowledge entities in the annotated drug development sub-graph is 6. Sort and integrate multiple task matching priorities to generate multiple matching priority integration sequences. Since M is divisible by 2, M / 2=3 matching priority integration sequences are obtained from multiple matching priority integration sequences. Reverse transform these 3 matching priority integration sequences to generate 3 reverse transformed matching priority integration sequences. Associate these 3 matching priority integration sequences and the 3 reverse transformed matching priority integration sequences to the knowledge entities annotated with 6 knowledge points, and one knowledge entity annotated with one knowledge point corresponds to one matching priority integration sequence. For example, for the drug chemist knowledge entity annotated as the knowledge point of "drug molecule synthesis", it is associated with a matching priority integration sequence, and its matching priority information is determined based on this sequence.

[0038] Step S140 : when determining the matching progress that meets the requirements for matching the target talent knowledge entity with the research and development task, determining the target research and development task corresponding to the target talent knowledge entity according to the matching priority information of the target talent knowledge entity.

[0039] In this embodiment, it is assumed that the target talent knowledge entity is the medicinal chemist mentioned above. The server continuously monitors the progress of each R&D task and the allocation of related resources in the entire R&D task management system to determine whether it meets the matching progress for matching R&D tasks for medicinal chemists. For example, when the scientific research platform has newly acquired a batch of advanced drug synthesis equipment and has completed the preliminary theoretical research work on the "new drug molecular structure design" task, it meets the matching progress for task matching for medicinal chemists.

[0040] The target R&D task corresponding to the medicinal chemist is determined based on the matching priority information of the medicinal chemist. Since the medicinal chemist has been determined to have a higher matching priority in the task of "new drug molecular structure design", "new drug molecular structure design" is determined as the target R&D task at this time.

[0041] Step S150: Associating the target R&D task with the target talent knowledge entity.

[0042] In this embodiment, after determining that "new drug molecular structure design" is the target R&D task and the target talent knowledge entity is a medicinal chemist, the R&D task of "new drug molecular structure design" is associated with the talent knowledge entity of the medicinal chemist in the R&D task management system within the scientific research platform. This association operation will be recorded in the system, including the detailed information of the task, the start time of the task, the expected completion time and other information are bound to the personal information, work schedule and the like of the medicinal chemist. For example, in the task allocation module of the task management system, it will be displayed that the person in charge of the "new drug molecular structure design" task is the medicinal chemist, and this task will also appear in the personal work task list of the medicinal chemist. At the same time, the task management system will allocate corresponding experimental equipment, research funds and other resources to the medicinal chemist according to the priority of the task and the allocation of resources to ensure the smooth implementation of the task.

[0043] Based on the above steps, the embodiment of the present application can efficiently achieve accurate matching of scientific research talents and R&D tasks. By acquiring and parsing multiple candidate R&D tasks and their matching priorities, combined with the research-related knowledge feature data between scientific research talent nodes, multiple scientific research talent knowledge sub-graphs are constructed. Furthermore, the scientific research talent nodes are prioritized according to the task matching priority, so that in the same scientific research talent knowledge sub-graph, scientific research talent nodes of different sequences have different matching priority information. In the matching process, according to the matching priority information of the target talent knowledge entity, the corresponding target R&D task is quickly determined, and the target R&D task is associated with the corresponding talent knowledge entity, thereby not only improving the matching efficiency of scientific research talents and R&D tasks, but also ensuring the accuracy and rationality of the matching, which helps to optimize the allocation of scientific research resources and promote the smooth progress of scientific research projects.

[0044] In a possible implementation, the step of determining research-related knowledge feature data between scientific research talent nodes includes:

[0045] Step A110, obtaining multi-source academic research data, wherein the multi-source academic research data includes academic paper database, scientific research project database, patent database, scholar social network data and institutional cooperation data.

[0046] Step A120, using natural language processing algorithm and entity recognition algorithm, to identify scientific research talent nodes of the multi-source academic research data, and generate a list of scientific research talent nodes, wherein each scientific research talent node in the list contains a unique identifier and basic information.

[0047] Step A130, for each scientific research talent node, extract the scientific research content characteristics of the scientific research achievement data of the scientific research talent node through text mining and semantic analysis algorithms, and the scientific research content characteristics include keyword feature vectors, theme feature vectors, and research field feature vectors.

[0048] Step A140, based on the cooperation information in the multi-source academic research data, construct a cooperation network diagram among scientific research talents, combine the scientific research content characteristics and the cooperation network diagram, calculate the research association knowledge characteristics between scientific research talent nodes, and generate a feature data matrix containing the research association intensity between scientific research talent nodes.

[0049] Step A150, optimizing the feature data matrix using linear discriminant analysis and recursive feature elimination algorithm to generate an optimized feature data matrix.

[0050] Step A160, using actual scientific research cooperation data as a verification set, verifies the optimized feature data matrix, and based on the verification results, performs advanced optimization on the optimized feature data matrix to generate research-related knowledge feature data between scientific research talent nodes.

[0051] And, step S120 includes:

[0052] Step S121, constructing a basic scientific research talent knowledge graph based on the research-related knowledge feature data between the scientific research talent nodes, each scientific research talent node is each knowledge entity in the basic scientific research talent knowledge graph, and there is research-related knowledge feature data between two scientific research talent nodes associated with an entity association link in the basic scientific research talent knowledge graph.

[0053] Step S122, dividing the basic scientific research talent knowledge graph according to the previously defined knowledge graph division strategy to generate multiple scientific research talent knowledge sub-graphs.

[0054] In this embodiment, multi-source academic research data are first obtained. These data come from a wide range of sources and are crucial to building scientific research talent relationships. The academic paper database is an important part of it, covering a large number of academic research results in different disciplines around the world. For example, in the fields of drug development and artificial intelligence algorithm research, the academic paper database stores various types of papers from basic theoretical research to applied practice exploration. The scientific research project database records in detail the situation of many scientific research projects within the scientific research platform and in cooperation with external parties, including project goals, participants, project cycles, and outputs. Taking the ongoing anti-cancer drug development project of the scientific research platform as an example, the scientific research project database contains detailed information on the project from drug molecule design to clinical trials at all stages. The patent database contains registered patent information in related fields, which is very critical for understanding the situation of scientific research talents in the protection of innovative achievements. In the field of drug development, new drug synthesis methods, active pharmaceutical ingredients, etc. may be patented, and these patent information are closely related to scientific research talents. Scholars' social network data reflects the social interaction between researchers, such as exchanges at international academic conferences, expressions of cooperation intentions, etc. Institutional cooperation data shows the cooperative relationship between the scientific research platform and other scientific research institutions, such as the cooperation between the scientific research platform and a well-known university on an artificial intelligence algorithm improvement project.

[0055] Next, the server uses natural language processing algorithms and entity recognition algorithms to identify scientific research talent nodes in multi-source academic research data, thereby generating a list of scientific research talent nodes. For papers in the academic paper database, the natural language processing algorithm analyzes and processes the paper text, and the entity recognition algorithm accurately identifies the names of scientific research talents and other information. For example, in a paper on the development of anti-cancer drugs, the algorithm can accurately identify the author's name, research institution, and other information, and use this information as the basic information of the scientific research talent node. Each scientific research talent node is assigned a unique identifier, such as an employee number or a unique code generated according to a specific rule, and also contains basic information such as name, education, and professional field. In this way, a comprehensive list of scientific research talent nodes is generated by processing a large amount of academic papers, scientific research project materials, patent documents, and scholars' social networks and institutional cooperation data.

[0056] For each scientific research talent node, the server extracts the scientific research content characteristics of the scientific research results data of the scientific research talent node through text mining and semantic analysis algorithms. Taking researchers in the field of drug research and development as an example, the server will deeply mine the papers they published, the reports of scientific research projects they participated in, and other materials. For the extraction of keyword feature vectors, the server will count the words frequently used by researchers in their achievements. For example, medicinal chemists may often use keywords such as "drug molecular structure", "chemical synthesis", and "active ingredients". These keywords and their frequency of occurrence constitute the keyword feature vector. In terms of topic feature vectors, it is determined according to the theme of the paper or the core content of the scientific research project. For example, a topic such as "R&D strategy for new anticancer drugs" can be extracted as a topic feature vector. The research field feature vector clarifies the major research field to which the researcher belongs, such as "pharmaceutical chemistry" and "clinical pharmacology". For artificial intelligence algorithm developers, their keywords may include "algorithm optimization" and "model training". The topic feature vector may be "improvement of image recognition algorithm based on deep learning", and the research field feature vector is "artificial intelligence algorithm research".

[0057] Based on the cooperation information in multi-source academic research data, the server constructs a cooperation network diagram between scientific research talents. In a drug development project, if a medicinal chemist and a pharmacologist jointly participate in a drug development project, there will be a link connecting these two scientific research talent nodes in the cooperation network diagram. Combining the previously extracted scientific research content features and the constructed cooperation network diagram, the server calculates the research-related knowledge features between scientific research talent nodes. For example, if two scientific research talent nodes have more connection links in the cooperation network diagram, and their scientific research content features have a high similarity (such as similar research fields, overlapping frequently used keywords, etc.), then the research association strength between them is high. Through this calculation between all scientific research talent nodes, a feature data matrix containing the research association strength between scientific research talent nodes is generated. Each element in this matrix represents the research association strength value between a pair of scientific research talent nodes.

[0058] Subsequently, the server uses linear discriminant analysis and recursive feature elimination algorithms to optimize the feature data matrix. There may be some features in the feature data matrix that do not contribute much to distinguishing the relationship between scientific research talents or are redundant. The linear discriminant analysis algorithm helps determine which features are most critical for distinguishing the relationship between scientific research talents by analyzing the differences between different categories. The recursive feature elimination algorithm starts with all features and gradually removes those that contribute the least to the model. For example, in the initial feature data matrix, there may be some low-frequency keywords or less important cooperation project information that have little actual correlation with scientific research talents. Through the optimization of these two algorithms, these unnecessary features can be removed to generate an optimized feature data matrix.

[0059] Then, the server uses the actual scientific research cooperation data as the verification set to verify the optimized feature data matrix. The actual scientific research cooperation data reflects the actual cooperation between scientific researchers. For example, in the actual drug development cooperation project within the scientific research platform, which scientific researchers actually cooperated and the degree of cooperation. According to the verification results, if the optimized feature data matrix shows that two scientific researchers have a high degree of correlation, but they rarely cooperate in the actual scientific research cooperation data, then the optimized feature data matrix needs to be adjusted. It may be necessary to re-examine the previous calculation process, adjust the weights of certain features, or reconsider the calculation methods of certain correlations. After such advanced optimization, the research-related knowledge feature data between scientific research talent nodes is finally generated.

[0060] Based on the research-related knowledge feature data between the above scientific research talent nodes, the server begins to build a basic scientific research talent knowledge graph. In this basic scientific research talent knowledge graph, each scientific research talent node becomes each knowledge entity in the basic scientific research talent knowledge graph. For example, in the basic scientific research talent knowledge graph related to drug development and artificial intelligence algorithm development, scientific research talent nodes such as medicinal chemists, pharmacologists, algorithm engineers, and data scientists are all knowledge entities. There is research-related knowledge feature data between the two scientific research talent nodes associated with the entity association link in the basic scientific research talent knowledge graph. For example, if two scientific research talents show a high correlation strength in the previously calculated research-related knowledge feature data, then the link between them in the basic scientific research talent knowledge graph represents this close correlation relationship.

[0061] Finally, the server divides the basic scientific research talent knowledge graph according to the previously defined knowledge graph division strategy to generate multiple scientific research talent knowledge sub-graphs. The knowledge graph division strategy may be based on a variety of factors. For example, if it is divided according to the research field, the scientific research talent nodes related to drug development will be divided into a scientific research talent knowledge sub-graph, and the scientific research talent nodes related to artificial intelligence algorithm development will be divided into another sub-graph. Or it can be divided according to the project team. If the scientific research platform has a dedicated anti-cancer drug development project team and an artificial intelligence algorithm optimization project team, then the relevant scientific research talent nodes can be divided into different scientific research talent knowledge sub-graphs according to the affiliation of team members. In this way, through careful division, multiple scientific research talent knowledge sub-graphs with specific meanings and close internal connections are obtained, which provides a data basis for subsequent operations such as precise matching of research and development tasks and scientific research talents.

[0062] In a possible implementation, step S130 includes:

[0063] Step S131, divide the knowledge entities in the a-th scientific research talent knowledge sub-graph into a first knowledge entity sequence and a second knowledge entity sequence, there is no duplication of knowledge entities in the first knowledge entity sequence and the second knowledge entity sequence, and the number of knowledge entities in the first knowledge entity sequence and the second knowledge entity sequence is the same.

[0064] Step S132: construct a unique mapping relationship between the first knowledge entity sequence and the second knowledge entity sequence, wherein a first knowledge entity in the first knowledge entity sequence and a second knowledge entity in the second knowledge entity sequence have a unique mapping relationship.

[0065] Step S133: sort and integrate the multiple task matching priorities to generate multiple matching priority integration sequences.

[0066] Step S134: determining a matching priority integration sequence corresponding to each first knowledge entity from the multiple matching priority integration sequences.

[0067] Step S135 : determining the matching priority information corresponding to each of the first knowledge entities according to the matching priority integration sequence corresponding to each of the first knowledge entities.

[0068] Step S136: reversely convert the matching priority information corresponding to each of the first knowledge entities to generate matching priority information of the second knowledge entities corresponding to each of the first knowledge entities.

[0069] In this embodiment, the operation is first performed on the ath scientific research talent knowledge sub-map, taking the scientific research talent knowledge sub-map related to drug development as an example. This sub-map contains numerous knowledge entities, such as various scientific research talent nodes such as medicinal chemists, pharmacologists, and clinical researchers. The server must distinguish these knowledge entities into a first knowledge entity sequence and a second knowledge entity sequence, and require that the knowledge entities in these two sequences do not have duplications and the number remains consistent. For example, if there are 10 knowledge entities in the sub-map, the server will divide 5 of the knowledge entities, such as medicinal chemists, clinical researchers, drug analysis experts, pharmacists, and toxicologists, into the first knowledge entity sequence; the remaining pharmacologists, drug metabolism researchers, drug formulation engineers, clinical pharmacologists, and drug quality control experts are divided into the second knowledge entity sequence. This division is based on a certain logical order or pre-set rules, and is intended to provide a reasonable structural basis for subsequent matching priority determination.

[0070] After completing the sequence division of knowledge entities, the server constructs a unique mapping relationship between the first knowledge entity sequence and the second knowledge entity sequence. The construction of this mapping relationship is based on the scientific research workflow and the intrinsic connection between knowledge entities. Taking drug development as an example, there is a unique mapping relationship between the medicinal chemist in the first knowledge entity sequence and the pharmacologist in the second knowledge entity sequence, because in the drug development process, after the medicinal chemist synthesizes a new drug molecule, the pharmacologist will study its pharmacological properties such as activity and toxicity; there is a mapping relationship between the clinical researcher in the first knowledge entity sequence and the drug metabolism researcher in the second knowledge entity sequence. The drug reaction data in the human body collected by the clinical researcher during the clinical trial provides research materials for the drug metabolism researcher. This unique mapping relationship accurately reflects the collaborative relationship between different types of scientific research talents in the entire R&D process.

[0071] Next, the server sorts and integrates the matching priorities of multiple tasks to generate multiple matching priority integration sequences. In the research and development task system of the scientific research platform, there are many different research and development tasks, and each task has a previously defined matching priority. For example, in a drug research and development project, the tasks include the design of new drug molecular structures, optimization of drug synthesis processes, and formulation of drug clinical trial plans. According to the strategic goals, market demand, and project progress of the scientific research platform, these tasks are given different matching priorities. The server comprehensively sorts and integrates the matching priorities of these tasks, and may generate a matching priority integration sequence such as "new drug molecular structure design > drug synthesis process optimization > drug clinical trial plan formulation" based on factors such as the importance, urgency, and resource requirements of the tasks. If there are tasks of artificial intelligence algorithm improvement projects at the same time, such as algorithm model architecture optimization, algorithm training data expansion and cleaning, and algorithm performance evaluation index system construction, they will also be comprehensively sorted and integrated with the matching priorities of drug research and development tasks to form multiple matching priority integration sequences covering different project tasks.

[0072] Then, the server determines the matching priority integration sequence corresponding to each first knowledge entity from multiple matching priority integration sequences. Taking the medicinal chemist in the first knowledge entity sequence as an example, since it plays a key role in the molecular structure design in the early stage of drug research and development, the server will find the matching priority integration sequence related to the task of "new drug molecular structure design" from the numerous matching priority integration sequences as its corresponding sequence. This process requires an in-depth analysis of the relevance of each task to each first knowledge entity in terms of knowledge, skills, and workflow. For clinical researchers, the server will determine the matching priority integration sequence related to the task of "drug clinical trial plan formulation" as its corresponding matching priority integration sequence, because the main work of clinical researchers is closely related to the formulation and implementation of clinical trial plans.

[0073] According to the matching priority integration sequence corresponding to each first knowledge entity, the server determines the matching priority information corresponding to each first knowledge entity. For example, for the matching priority integration sequence corresponding to the medicinal chemist, if the task of "new drug molecular structure design" is at the highest priority in the sequence, then the matching priority information of the medicinal chemist indicates that it has the highest matching degree in the research and development work related to this task. For clinical researchers, if the task of "drug clinical trial plan formulation" is at a higher position in its corresponding matching priority integration sequence, then the clinical researcher has a higher matching priority when performing research and development work related to this task. This matching priority information clarifies the relative importance and adaptability of each first knowledge entity in the entire research and development task system.

[0074] Finally, the server performs reverse conversion on the matching priority information corresponding to each first knowledge entity, and generates matching priority information of the second knowledge entity corresponding to each first knowledge entity. Taking the mapping relationship between the medicinal chemist and the pharmacologist mentioned above as an example, according to the matching priority information of the medicinal chemist, if the medicinal chemist has a high matching priority in the task of "new drug molecular structure design", then when performing reverse conversion, for the pharmacologist, the server will determine the matching priority in the pharmacological research (such as drug activity testing, etc.) tasks related to the subsequent task according to the logical relationship of the entire R&D process. For the mapping relationship between clinical researchers and drug metabolism researchers, if the clinical researcher has a specific matching priority in the task of "drug clinical trial plan formulation", then the matching priority of the drug metabolism researcher in the tasks related to drug metabolism research after clinical trial data collection will be reversely converted and determined according to the matching priority information of the clinical researcher. Through such reverse conversion operations, the knowledge entities in the first knowledge entity sequence and the second knowledge entity sequence can form an interrelated and coordinated priority system when matching R&D tasks, thereby providing a data basis for the reasonable allocation of R&D tasks to each scientific research talent node in the entire scientific research talent knowledge sub-map.

[0075] In a possible implementation, step S131 includes:

[0076] Step S1311, obtaining the number of knowledge entities in the a-th scientific research talent knowledge sub-graph, a=1, 2, .., X, X is the number of scientific research talent knowledge sub-graphs, and X is a positive integer not less than 2.

[0077] Step S1312, when the number of the knowledge entities is divisible by 2, the knowledge entities in the a-th scientific research talent knowledge sub-graph are all the first knowledge entity sequence and the second knowledge entity sequence.

[0078] Step S1313, when the number of the knowledge entities cannot be divided by 2, a target knowledge entity is randomly extracted from the ath scientific research talent knowledge sub-graph.

[0079] Step S1314, dividing the other knowledge entities except the target knowledge entity in the a-th scientific research talent knowledge sub-graph into a first knowledge entity sequence and a second knowledge entity sequence on average.

[0080] In this embodiment, it is first necessary to obtain the number of knowledge entities in the ath scientific research talent knowledge sub-graph, where a is 1, 2, and so on up to X, where X is the number of scientific research talent knowledge sub-graphs and is a positive integer not less than 2. Taking the scientific research talent knowledge sub-graph related to drug development in the scientific research platform as an example, assume that there are 3 scientific research talent knowledge sub-graphs, corresponding to different drug development stages or different types of drug development projects. When the server starts processing the first scientific research talent knowledge sub-graph, it will query and obtain the number of knowledge entities in this sub-graph.

[0081] When the number of knowledge entities is divisible by 2, the server will evenly divide the knowledge entities in the a-th scientific research talent knowledge subgraph into the first knowledge entity sequence and the second knowledge entity sequence. For example, in a scientific research talent knowledge subgraph for drug development, there are 10 knowledge entities, which cover scientific research talents in different professional fields, from drug chemistry researchers, pharmacologists to clinical researchers. The server will divide these 10 knowledge entities evenly according to the rules. Specifically, the five knowledge entities of drug chemists, pharmacologists, drug analysis experts, pharmacists, and toxicologists may be divided into the first knowledge entity sequence; and the other five knowledge entities of clinical researchers, drug metabolism researchers, drug formulation engineers, clinical pharmacologists, and drug quality control experts may be divided into the second knowledge entity sequence. This division method ensures the quantitative balance of the two sequences and lays the foundation for the subsequent construction of a reasonable matching priority relationship.

[0082] However, when the number of knowledge entities in the ath scientific research talent knowledge subgraph cannot be divided by 2, the server will adopt another division strategy. For example, in another scientific research talent knowledge subgraph related to a special drug development project, there are 7 knowledge entities. The server will arbitrarily extract a target knowledge entity from this subgraph, assuming that this target knowledge entity is a pharmaceutical formulation engineer. Then, the server will evenly divide the other 6 knowledge entities in this subgraph except the pharmaceutical formulation engineer into the first knowledge entity sequence and the second knowledge entity sequence. That is, the three knowledge entities of pharmaceutical chemists, pharmacologists, and drug analysis experts are divided into the first knowledge entity sequence; the three knowledge entities of clinical researchers, drug metabolism researchers, and drug quality control experts are divided into the second knowledge entity sequence. Although this division method is not completely equal in quantity, by reasonably selecting the target knowledge entities and dividing them evenly, two relatively reasonable knowledge entity sequences can still be constructed, thereby meeting the subsequent needs for determining the priority information of scientific research talent node matching.

[0083] This method of distinguishing the first knowledge entity sequence from the second knowledge entity sequence is to establish an orderly structure in the scientific research talent knowledge sub-graph, so as to further determine the matching priority information of each scientific research talent node according to the task matching priority. Regardless of whether it is divisible by 2 or not, the operation of the server is based on ensuring the rigor and effectiveness of the subsequent processing logic, ensuring that an accurate and reasonable matching relationship can be established between scientific research talents and R&D tasks, thereby improving the management efficiency of the entire scientific research project and the output quality of R&D results.

[0084] In the actual scientific research management process, this distinction helps to integrate the resources and capabilities of different types of scientific research talents. For example, in different links of drug research and development, scientific research talents in the first knowledge entity sequence may play a leading role in the early basic research and drug development stages, while scientific research talents in the second knowledge entity sequence play an important role in subsequent clinical research, drug production and quality control. Through this clear division, the server can better coordinate the role positioning and task allocation of different scientific research talents in the entire research and development process, so that the research and development work in each link can be carried out in an orderly and efficient manner. At the same time, this division method is also convenient for the classified management and evaluation of scientific research talents. According to the characteristics and task requirements of scientific research talents in different sequences, targeted training plans and performance evaluation standards are formulated to promote the growth and development of scientific research talents, thereby promoting the improvement of the scientific research strength of the entire scientific research platform.

[0085] In a possible implementation, step S135 includes:

[0086] Step S1351, obtaining Y task matching priorities included in the matching priority integration sequence corresponding to the bth first knowledge entity, where Y is a positive integer not less than 2, b=1, 2, .., Z, where Z is the number of first knowledge entities, and Z is a positive integer not less than 2.

[0087] Step S1352: output the first Y task matching priorities in the matching priority integration sequence corresponding to the bth first knowledge entity as the matching priorities for associating the first Y candidate R&D tasks to the bth first knowledge entity.

[0088] In this embodiment, first obtain the Y task matching priorities contained in the matching priority integration sequence corresponding to the b-th first knowledge entity, where Y is a positive integer not less than 2, b takes a value from 1 to Z, and Z is the number of first knowledge entities and is not less than 2. Taking the drug development project of the scientific research platform as an example, assuming that in the knowledge sub-graph of scientific research talents related to drug development, there are 5 first knowledge entities in the first knowledge entity sequence (i.e., Z=5), and the second first knowledge entity (b=2) is used as an example for explanation. This first knowledge entity may be a medicinal chemist, and the corresponding matching priority integration sequence contains multiple task matching priorities (Y), which are derived from the research and development task system of the entire scientific research platform, including tasks in different projects such as drug development and artificial intelligence algorithm improvement. For example, this matching priority integration sequence may include matching priorities for new drug molecular structure design tasks, drug synthesis process optimization tasks, artificial intelligence algorithm model architecture optimization tasks, algorithm training data expansion and cleaning tasks, etc.

[0089] Then, the server outputs the first Y task matching priorities in the matching priority integration sequence corresponding to the b-th first knowledge entity (here, the medicinal chemist) as the matching priority for associating the first Y candidate R&D tasks to the b-th first knowledge entity. Assuming Y=3, in the matching priority integration sequence corresponding to the medicinal chemist, the first three task matching priorities correspond to the new drug molecular structure design task, the drug synthesis process optimization task, and the artificial intelligence algorithm model architecture optimization task (here is just an example to illustrate that the integration sequence may include tasks of different projects). Then the server will determine the matching priorities of these three tasks for the first knowledge entity of the medicinal chemist. For the medicinal chemist, the new drug molecular structure design task is at the front position in its matching priority integration sequence, which means that among all the candidate R&D tasks, the new drug molecular structure design task has the highest matching degree with the medicinal chemist, and this task should be assigned to the medicinal chemist as a priority when allocating tasks. The drug synthesis process optimization task is in second place, and its matching priority is second. Although the artificial intelligence algorithm model architecture optimization task is also among the first Y tasks, its matching priority is relatively low because it is at the back of the matching priority integration sequence of drug chemists.

[0090] This method of determining the matching priority information of the first knowledge entity helps to accurately associate the candidate R&D tasks with the first knowledge entity. In actual scientific research management, this operation allows the scientific research platform to reasonably arrange R&D tasks according to the professional capabilities and characteristics of scientific research talents. For example, for drug research and development projects, the scientific research platform can ensure that key scientific research talents such as medicinal chemists first undertake tasks that best match their professional skills and knowledge background, thereby improving R&D efficiency. At the same time, since the tasks of different projects are reflected in the matching priority integration sequence, it is also conducive to cross-project resource allocation and task coordination. If the scientific research platform needs to invest more resources in key tasks in drug research and development at a certain period, such as the task of designing the molecular structure of new drugs, through the determination of this matching priority, the high priority of medicinal chemists on this task can be clarified, so as to reasonably allocate human, material and financial resources. Moreover, this method also provides a basis for subsequent scientific research talent management. For example, when conducting performance appraisal or career development planning for medicinal chemists, they can be evaluated based on their performance on high matching priority tasks, and their roles and task allocation in future projects can be planned based on the priority relationship between these tasks and other tasks.

[0091] In the entire scientific research system, this method of determining the first knowledge entity matching priority information based on the matching priority integration sequence is a key link. It not only involves the precise matching of individual scientific research talents with specific research and development tasks, but also relates to the resource allocation, schedule arrangement and final output of the entire scientific research project. Whether it is for traditional projects such as drug development or emerging projects such as artificial intelligence algorithm improvement, this method can ensure that scientific research talents undertake the most suitable tasks in the most suitable positions, thereby promoting the continuous development of the scientific research platform.

[0092] In a possible implementation, step S136 includes:

[0093] Step S1361: According to the matching priority information corresponding to the bth first knowledge entity, the first Y candidate R&D tasks are arranged in ascending order of matching priority to generate sorted Y candidate R&D tasks.

[0094] Step S1362 , arranging the Y task matching priorities in descending order to generate sorted Y task matching priorities.

[0095] Step S1363: output the sorted first Y task matching priorities as the matching priorities for associating the sorted first Y candidate R&D tasks to the bth second knowledge entity corresponding to the bth first knowledge entity.

[0096] In this embodiment, taking the knowledge sub-graph of scientific research talents related to drug development as an example, it is assumed that there is a first knowledge entity called a medicinal chemist, and its corresponding matching priority information indicates the priority order of several candidate research and development tasks. When operating according to the matching priority information corresponding to the bth (here it is assumed that b is the identifier of the first knowledge entity corresponding to the medicinal chemist) first knowledge entity, the server will first arrange the first Y candidate research and development tasks in ascending order according to the matching priority, thereby generating the sorted Y candidate research and development tasks. For example, if the initial candidate research and development tasks include new drug molecular structure design, drug synthesis process optimization, drug clinical trial program formulation, etc., and according to the matching priority information of the medicinal chemist, the new drug molecular structure design task originally has the highest priority, the drug synthesis process optimization is second, and the drug clinical trial program formulation is the lowest, then after the ascending order, it becomes the order of drug clinical trial program formulation, drug synthesis process optimization, and new drug molecular structure design.

[0097] Next, the server arranges the Y task matching priorities in descending order to generate the sorted Y task matching priorities. Since the tasks were previously arranged in ascending order, the task matching priorities are now arranged in reverse order. For example, for the three tasks mentioned above, the task matching priority value of the new drug molecular structure design task was originally the highest under the matching priority of the medicinal chemist (assuming it is 3), followed by the drug synthesis process optimization (assuming it is 2), and the drug clinical trial plan formulation is the lowest (assuming it is 1). Then, after arranging the task matching priorities in descending order, it becomes the order of 1, 2, 3, except that the corresponding tasks become the tasks after ascending order, that is, the task matching priority corresponding to the drug clinical trial plan formulation becomes 3, the corresponding task matching priority corresponding to the drug synthesis process optimization becomes 2, and the corresponding task matching priority corresponding to the new drug molecular structure design becomes 1.

[0098] Finally, the server outputs the sorted first Y task matching priorities as the matching priorities of associating the sorted first Y candidate R&D tasks to the bth second knowledge entity corresponding to the bth first knowledge entity. Continuing with the above example, for the second knowledge entity corresponding to the medicinal chemist (such as a pharmacologist), the sorted task matching priorities determine the matching priorities of these candidate R&D tasks for the pharmacologist. This means that after the medicinal chemist completes the drug molecular structure design, the pharmacological research tasks related to the formulation of the drug clinical trial plan have the highest matching priority when associated with him in the subsequent R&D process, because this task has the highest priority value in the reverse conversion matching priority, while the pharmacological research tasks related to the molecular structure design of new drugs (such as the preliminary pharmacological properties estimation of new structure molecules, etc.) have the lowest matching priority. This reverse conversion matching priority information enables the second knowledge entity to reversely determine the priority of its own tasks according to the task priority of the first knowledge entity when cooperating with the first knowledge entity in R&D tasks, ensuring the coherence and logic of the entire R&D process, making the collaboration of different scientific research talents in R&D tasks more orderly and efficient, and helping to improve the overall efficiency and success rate of the entire drug R&D project. In the complex environment where the scientific research platform involves multiple scientific research projects and numerous scientific research talents, this mechanism of generating the second knowledge entity matching priority information based on the reverse conversion of the first knowledge entity matching priority information provides an effective technical means for the rational arrangement of scientific research tasks and the optimization of resource allocation.

[0099] In a possible implementation manner, the step S130 may further include:

[0100] Step B110, annotating the multiple scientific research talent knowledge sub-graphs to generate multiple annotated scientific research talent knowledge sub-graphs, in which the annotated knowledge points of the two knowledge entities of the first entity association link Guanlan are different.

[0101] Step B120, obtain the number M of knowledge points annotated by knowledge entities in the a-th annotated scientific research talent knowledge sub-graph, where a=1, 2, .., X, where X is the number of scientific research talent knowledge sub-graphs, and X is a positive integer not less than 2.

[0102] Step B130: sort and integrate the multiple task matching priorities to generate multiple matching priority integration sequences.

[0103] Step B140: when M is divisible by 2, obtain M / 2 matching priority integration sequences from the multiple matching priority integration sequences.

[0104] Step B150, reversely convert the M / 2 matching priority integration sequences to generate M / 2 reversely converted matching priority integration sequences.

[0105] Step B160, associating the M / 2 matching priority integration sequences and the M / 2 matching priority integration sequences after reverse conversion to the knowledge entities of the M labeled knowledge points, one knowledge entity labeled with a knowledge point corresponds to one matching priority integration sequence.

[0106] Step B170, determining the matching priority information of each scientific research talent node based on the matching priority integration sequence corresponding to each knowledge entity.

[0107] In a possible implementation, step B110 includes:

[0108] Step B111, obtaining a set of annotated basic information for annotating the scientific research talent knowledge sub-graph from multiple predetermined data sources, wherein the data sources include a subject classification system library, a terminology standard library in the scientific research field, and an industry general research direction guide library.

[0109] Step B112, for the research and development data of the knowledge entities in each scientific research talent knowledge sub-graph, preliminary annotation is performed based on the annotation basic information in the annotation basic information set to generate a scientific research talent knowledge sub-graph set after preliminary annotation.

[0110] Step B113, based on the set of scientific research talent knowledge sub-graphs after preliminary annotation and the set of annotation basic information, determine the annotation basis of the entity association relationship, generate a relationship annotation basis set, the relationship annotation basis set includes all relationship annotation rules constructed based on knowledge entity annotation and annotation basic information, specifically analyze the logical relationship between different knowledge entity annotations, combine the research direction information and subject association rules in the annotation basic information, and construct a relationship annotation rule system.

[0111] Step B114, according to the rules in the relationship annotation basis set, annotate the entity association relationship in each of the scientific research talent knowledge sub-graphs. Specifically, for each pair of knowledge entities with an associated relationship, determine the annotation content of the associated relationship according to the corresponding entity annotation information and the pre-constructed relationship annotation rules, thereby generating a set of scientific research talent knowledge sub-graphs after relationship annotation as multiple annotated scientific research talent knowledge sub-graphs.

[0112] In this embodiment, first, multiple scientific research talent knowledge sub-graphs are annotated to generate multiple annotated scientific research talent knowledge sub-graphs. In this process, the annotated knowledge points of two knowledge entities associated with an entity association link in the annotated scientific research talent knowledge sub-graph are different. The server obtains a set of annotated basic information for annotating scientific research talent knowledge sub-graphs from multiple predetermined data sources, including a subject classification system library, a terminology standard library in the scientific research field, and an industry-wide research direction guide library. For example, in the scientific research talent knowledge sub-graph related to drug research and development, the subject classification system library provides subject classification information such as medicinal chemistry, pharmacology, and pharmacy; the terminology standard library in the scientific research field contains professional terminology standards such as "active drug ingredients", "drug metabolism", and "clinical trials"; the industry-wide research direction guide library gives the current hot research directions in the drug research and development industry, such as new anti-cancer drug development, personalized drug therapy, and other information.

[0113] For the research and development data of each knowledge entity in the scientific research talent knowledge sub-graph, the server performs preliminary annotation based on the annotation basic information in the annotation basic information set, and generates a set of scientific research talent knowledge sub-graphs after preliminary annotation. Taking the scientific researchers in drug research and development as an example, for the knowledge entity of medicinal chemist, according to the chemical synthesis, drug molecular structure design and other contents involved in its research and development data, combined with the classification of medicinal chemistry disciplines and relevant terminology standards in the annotation basic information, it is preliminarily annotated as a knowledge entity related to medicinal chemistry, involving research and development directions such as drug molecular structure design. For the knowledge entity of pharmacologist, according to its research and development data in drug activity testing, drug mechanism research, etc., it is preliminarily annotated according to the pharmacology-related standards in the annotation basic information.

[0114] Based on the preliminary labeled scientific research talent knowledge sub-graph set and the labeled basic information set, the server determines the labeling basis of the entity association relationship and generates a relationship labeling basis set. This set contains all relationship labeling rules constructed based on the knowledge entity labeling and the labeling basic information. For example, in drug research and development, if a medicinal chemist knowledge entity is labeled as focusing on the molecular structure design of new anticancer drugs, and a pharmacologist knowledge entity is labeled as engaged in anticancer drug activity testing research, then according to the discipline association rules (the order and collaborative relationship between medicinal chemistry and pharmacology in the drug research and development process) and the research direction information in the labeling basic information (both for anticancer drugs), the relationship labeling rules between the two are constructed, such as "the molecular structure designed by the medicinal chemist provides an object for the activity test of the pharmacologist". Through the specific analysis of the logical relationship between the labeling of different knowledge entities, combined with various information in the labeling basic information, a complete relationship labeling rule system is constructed.

[0115] According to the rules in the relationship annotation collection, the server annotates the entity associations in each scientific research talent knowledge sub-graph. For each pair of knowledge entities with an association relationship, the annotation content of the association relationship is determined according to the corresponding entity annotation information and the pre-constructed relationship annotation rules, thereby generating a set of scientific research talent knowledge sub-graphs with annotated relationships, as multiple annotated scientific research talent knowledge sub-graphs. For example, in a scientific research talent knowledge sub-graph of a drug development project, there is an association relationship between a medicinal chemist and a pharmacist. The annotation information of the medicinal chemist is "new drug molecular structure design", and the annotation information of the pharmacist is "making the drug into a suitable dosage form". According to the relationship annotation rule "the molecular structure designed by the medicinal chemist is the basis for the pharmacist to make the dosage form", the association between them is annotated.

[0116] Next, the server obtains the number M of knowledge entity annotation knowledge points in the a-th annotated scientific research talent knowledge sub-graph, where a is 1, 2, and so on up to X, where X is the number of scientific research talent knowledge sub-graphs and is not less than 2. For example, in a certain annotated scientific research talent knowledge sub-graph related to drug development, the number M of knowledge entity annotation knowledge points is determined to be 6 through statistics.

[0117] At the same time, the server sorts and integrates the matching priorities of multiple tasks to generate multiple matching priority integration sequences. In the research and development task system of the scientific research platform, there are various tasks, such as new drug molecular structure design tasks, drug synthesis process optimization tasks, drug clinical trial program formulation tasks, etc. in drug research and development, and algorithm model architecture optimization tasks, algorithm training data expansion and cleaning tasks, etc. in artificial intelligence algorithm improvement projects. According to the strategic goals, project requirements, resource allocation and other factors of the scientific research platform, the matching priorities of these tasks are sorted and integrated. For example, the matching priority integration sequence that may be generated is new drug molecular structure design task>drug synthesis process optimization task>drug clinical trial program formulation task>algorithm model architecture optimization task>algorithm training data expansion and cleaning task, etc.

[0118] When M is divisible by 2, such as the case where M is 6 mentioned above, the server obtains M / 2 (i.e. 3) matching priority integration sequences from multiple matching priority integration sequences. Assume that the matching priority integration sequences obtained are corresponding to the three tasks most relevant to drug development (new drug molecular structure design task, drug synthesis process optimization task, and drug clinical trial plan formulation task).

[0119] Then, the server performs reverse conversion on the M / 2 matching priority integration sequences to generate M / 2 reverse converted matching priority integration sequences. Taking the three matching priority integration sequences obtained previously as an example, for the matching priority integration sequence of the new drug molecular structure design task, if the task priorities were originally arranged in a certain order, the reverse conversion will rearrange the task priorities in the opposite order.

[0120] Afterwards, the server associates the M / 2 matching priority integration sequences and the M / 2 reversely converted matching priority integration sequences to the M knowledge entities annotated with knowledge points, and one knowledge entity annotated with a knowledge point corresponds to one matching priority integration sequence. For example, for a knowledge entity annotated as "drug molecular structure design related", it is associated with a converted or unconverted matching priority integration sequence, and for a knowledge entity annotated as "drug activity test related", it is associated with another matching priority integration sequence, and so on.

[0121] Finally, the matching priority information of each scientific research talent node is determined based on the matching priority integration sequence corresponding to each knowledge entity. Taking scientific research talents in drug research and development as an example, if the task of designing the molecular structure of a new drug is at a higher priority position in the matching priority integration sequence corresponding to a medicinal chemist knowledge entity, then this medicinal chemist has a higher matching priority information for this task, indicating that he has a higher adaptability when performing research and development work related to this task. This method determines accurate matching priority information for each scientific research talent node in the scientific research talent knowledge subgraph through a series of rigorous annotation, association and determination steps, which helps to achieve reasonable task allocation and optimal resource allocation in the scientific research management of the scientific research platform, and improve the operating efficiency of the entire scientific research system and the output quality of research and development results.

[0122] In a possible implementation, determining the matching priority information of each scientific research talent node in the plurality of scientific research talent knowledge sub-graphs may also include:

[0123] When M is not divisible by 2, (M+1) / 2 matching priority integration sequences are obtained from the multiple matching priority integration sequences.

[0124] Reverse conversion is performed on (M-1) / 2 matching priority integration sequences among the (M+1) / 2 matching priority integration sequences to generate (M-1) / 2 matching priority integration sequences after reverse conversion.

[0125] The (M+1) / 2 matching priority integration sequences and the (M-1) / 2 matching priority integration sequences after reverse conversion are associated with the knowledge entities of the M labeled knowledge points, and one knowledge entity labeled with a knowledge point corresponds to one matching priority integration sequence.

[0126] The matching priority information of each scientific research talent node is determined based on the matching priority integration sequence corresponding to each knowledge entity.

[0127] In this embodiment, taking the scientific research talent knowledge sub-graph related to drug research and development as an example, it is assumed that after the previous annotation processing, the number of knowledge points M of knowledge entity annotations in a certain annotated scientific research talent knowledge sub-graph is 5. The server first obtains (M+1) / 2 matching priority integration sequences from multiple matching priority integration sequences. In this case, (M+1) / 2=(5+1) / 2=3 matching priority integration sequences are obtained. These matching priority integration sequences contain the priority information of various R&D tasks in the scientific research platform. For example, in the field of drug research and development, there may be priority orders for new drug molecular structure design tasks, drug synthesis process optimization tasks, and drug clinical trial program formulation tasks. In the artificial intelligence algorithm improvement project, there are priority orders for algorithm model architecture optimization tasks, algorithm training data expansion and cleaning tasks, etc.

[0128] Next, the server reversely converts (M-1) / 2 matching priority integration sequences among the obtained (M+1) / 2 matching priority integration sequences to generate (M-1) / 2 reversely converted matching priority integration sequences. For the case where M is 5, (M-1) / 2=(5-1) / 2=2 matching priority integration sequences need to be reversely converted. Assuming that the three matching priority integration sequences obtained correspond to the new drug molecular structure design task, the drug synthesis process optimization task, and the drug clinical trial plan formulation task, respectively, then two of them (such as the matching priority integration sequences corresponding to the drug synthesis process optimization task and the drug clinical trial plan formulation task) are reversely converted. If the original order of the drug synthesis process optimization task in its matching priority integration sequence is before the drug clinical trial plan formulation task, after the reverse conversion, the priority of the drug clinical trial plan formulation task becomes before the drug synthesis process optimization task in the new sequence.

[0129] Then, the server associates (M+1) / 2 matching priority integration sequences and (M-1) / 2 matching priority integration sequences after reverse conversion to the knowledge entities of M annotated knowledge points, and one knowledge entity of annotated knowledge point corresponds to one matching priority integration sequence. For example, for the knowledge entity annotated as "knowledge related to drug molecular structure design", it is associated with the matching priority integration sequence corresponding to the new drug molecular structure design task; for the knowledge entity annotated as "knowledge related to drug synthesis process", it is associated with the matching priority integration sequence corresponding to the drug synthesis process optimization task after reverse conversion; for the knowledge entity annotated as "knowledge related to drug clinical trial", it is associated with another matching priority integration sequence corresponding to the drug clinical trial plan formulation task after reverse conversion, and so on. The knowledge entities of these five annotated knowledge points are associated with the corresponding matching priority integration sequences in turn.

[0130] Finally, the matching priority information of each scientific research talent node is determined according to the matching priority integration sequence corresponding to each knowledge entity. Taking the scientific research talent in drug research and development as an example, if the matching priority integration sequence corresponding to a medicinal chemist knowledge entity is the sequence associated with the new drug molecular structure design task, and the new drug molecular structure design task is in a higher priority position in this sequence, then this medicinal chemist has a higher matching priority information on this task. This means that when performing task allocation and resource allocation, this medicinal chemist has a higher adaptability in the research and development work related to the new drug molecular structure design. Similarly, for other scientific research talent nodes, such as pharmacologists and pharmacists, according to the priority of the tasks in the matching priority integration sequence they are associated with, their matching priority information on their respective related tasks is determined. This determination method helps to accurately match research and development tasks with scientific research talents in the scientific research management of the scientific research platform, so that scientific research talents can carry out their work in an orderly manner according to their professional capabilities and task priorities, thereby improving the efficiency of the entire scientific research system and the output quality of research and development results.

[0131] In a possible implementation, the method further includes:

[0132] Step C110, obtaining the R&D task transfer instruction generated by the first interactive port corresponding to the first scientific research talent node, wherein the R&D task transfer instruction includes a characteristic tag of the first R&D task to be transferred and a characteristic tag of the second scientific research talent node.

[0133] Step C120, based on the characteristic tag of the first R&D task and the characteristic tag of the second scientific research talent node, generates R&D task transfer push information to the second interactive port corresponding to the second scientific research talent node, and the R&D task transfer push information includes the characteristic tag of the first scientific research talent node and the characteristic tag of the first R&D task.

[0134] Step C130, when it is detected that the second scientific research talent node obtains the first push information of the first research and development task, adjust the research and development task holding information of the first scientific research talent node and the second scientific research talent node.

[0135] In this embodiment, first, the server obtains the R&D task transfer instruction generated by the first interactive port corresponding to the first scientific research talent node. Taking the drug research and development project as an example, the first scientific research talent node may be a medicinal chemist. In his work process, due to some reasons (such as excessive workload, encountering technical bottlenecks, or finding that another scientific research talent is more suitable for the task, etc.), he generates a R&D task transfer instruction through his first interactive port (which can be a software interface or interface specifically used for scientific researchers to interact with scientific research management systems). This instruction contains the characteristic label of the first R&D task transferred and the characteristic label of the second scientific research talent node. Assuming that the first R&D task is the "new anticancer drug molecular structure optimization" task, its characteristic label may include information such as the project name to which the task belongs, the technical field of the task (such as molecular structure optimization in medicinal chemistry), and the current stage of the task (such as the optimization stage after preliminary design). The second scientific research talent node is a senior medicinal chemist or a computational chemist with special skills, and its characteristic label may include his professional skills (such as being good at quantum chemical calculation-assisted molecular structure optimization), his team or department, and his work experience level.

[0136] Next, based on the characteristic tag of the first R&D task and the characteristic tag of the second scientific research talent node, the server generates a push message for the transfer of the R&D task to the second interactive port corresponding to the second scientific research talent node. The server will analyze the nature, requirements and current status of the task in detail according to the characteristic tag of the first R&D task, and at the same time, combine the characteristic tag of the second scientific research talent node to determine the potential adaptability of the task to the second scientific research talent node. Then, the second interactive port corresponding to the second scientific research talent node (similar to the interactive port of the first scientific research talent node, which is a channel specially provided for the researcher to interact with the system) generates a push message for the transfer of the R&D task. This push message contains the characteristic tag of the first scientific research talent node and the characteristic tag of the first R&D task. For example, the push message will inform the second scientific research talent node that the first scientific research talent node (the medicinal chemist) wants to transfer the task of "optimization of molecular structure of new anticancer drugs" due to some of its own circumstances, and it will also include the relevant characteristics of the first scientific research talent node, such as his previous work results in the project, his professional background, etc., so that the second scientific research talent node can fully understand the source and relevant background of the task.

[0137] Finally, after detecting that the second research talent node obtains the first push information of the first research and development task, the server adjusts the research and development task holding information of the first research talent node and the second research talent node. When the second research talent node obtains the push information through its interactive port, the server will monitor this event. Then, the server will update the task list of the first research talent node in the task management module of the scientific research management system, mark the "new anticancer drug molecular structure optimization" task as transferred or to be released, and update the task list of the second research talent node at the same time, marking the task as newly received or to be confirmed. This adjustment ensures the accuracy of the research and development task holding information in the scientific research management system, so that the task allocation of each research talent node can be reflected in the system in a timely and accurate manner, which is convenient for subsequent resource allocation, progress tracking and performance evaluation and other management operations. For example, in terms of resource allocation, the system may reallocate the experimental equipment and computing resources related to the task according to the situation of the second research talent node taking over the task; in terms of progress tracking, the expected completion time and intermediate checkpoints of the task can be re-planned according to the new task holder; in performance evaluation, the conversion of the responsible subject can be clarified and the work results of the new task bearer can be accurately evaluated.

[0138] Figure 2 The hardware structure of the system 100 for intelligent matching of scientific research talents and R&D tasks based on knowledge graphs provided in an embodiment of the present invention for realizing the above-mentioned intelligent matching method of scientific research talents and R&D tasks based on knowledge graphs is shown, as shown in FIG. Figure 2 As shown, the knowledge graph-based intelligent matching system 100 for scientific research talents and R&D tasks may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0139] The machine-readable storage medium 120 may store data and / or instructions. In some embodiments, the machine-readable storage medium 120 may store data obtained from an external terminal. In some embodiments, the machine-readable storage medium 120 may store data and / or instructions that the knowledge graph-based scientific research talent and R&D task intelligent matching system 100 uses to execute or use to complete the exemplary method described in the present invention.

[0140] During the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the intelligent matching method of scientific research talents and R&D tasks based on the knowledge graph in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0141] The specific implementation process of the processor 110 can refer to the various method embodiments executed by the above-mentioned knowledge graph-based scientific research talent and R&D task intelligent matching system 100. The implementation principles and technical effects are similar, and this embodiment will not be repeated here.

[0142] In addition, an embodiment of the present invention also provides a readable storage medium, in which computer executable instructions are preset. When the processor executes the computer executable instructions, the above-mentioned method for intelligent matching of scientific research talents and R&D tasks based on knowledge graph is implemented.

[0143] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, various features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A method for intelligently matching scientific research talents and R&D tasks based on knowledge graph, characterized in that: The method comprises: Obtain matching priorities of multiple candidate R&D tasks and multiple previously defined tasks; Acquire multiple scientific research talent nodes and research-related knowledge feature data between the scientific research talent nodes for matching R&D tasks, and determine multiple scientific research talent knowledge sub-graphs based on the research-related knowledge feature data; Determine the matching priority information of each scientific research talent node in the multiple scientific research talent knowledge sub-graphs according to the multiple task matching priorities, wherein the matching priority information of scientific research talent nodes in different sequences in the same scientific research talent knowledge sub-graph is different; When determining a matching progress that is suitable for matching a research and development task for a target talent knowledge entity, determining a target research and development task corresponding to the target talent knowledge entity according to matching priority information of the target talent knowledge entity; The target R&D task is associated with the target talent knowledge entity.

2. The method for intelligently matching scientific research talents and R&D tasks based on knowledge graph according to claim 1, characterized in that: The steps of determining the research-related knowledge feature data between scientific research talent nodes include: Acquire multi-source academic research data, including academic paper database, scientific research project database, patent database, scholar social network data and institutional cooperation data; Using a natural language processing algorithm and an entity recognition algorithm, the multi-source academic research data is used to identify scientific research talent nodes, and a scientific research talent node list is generated, wherein each scientific research talent node in the scientific research talent node list contains a unique identifier and basic information; For each scientific research talent node, the scientific research content features of the scientific research achievement data of the scientific research talent node are extracted through text mining and semantic analysis algorithms, and the scientific research content features include keyword feature vectors, theme feature vectors, and research field feature vectors; Based on the cooperation information in the multi-source academic research data, a cooperation network diagram among scientific research talents is constructed, and the research association knowledge characteristics among scientific research talent nodes are calculated by combining the scientific research content characteristics and the cooperation network diagram to generate a feature data matrix containing the research association strength among scientific research talent nodes; The feature data matrix is ​​optimized by using a linear discriminant analysis and a recursive feature elimination algorithm to generate an optimized feature data matrix; Using actual scientific research cooperation data as a verification set, the optimized feature data matrix is ​​verified, and according to the verification result, the optimized feature data matrix is ​​further optimized to generate research-related knowledge feature data between scientific research talent nodes; And, the determining of a plurality of scientific research talent knowledge sub-graphs based on the research-related knowledge feature data includes: A basic scientific research talent knowledge graph is constructed based on the research-related knowledge feature data between the scientific research talent nodes, each scientific research talent node is each knowledge entity in the basic scientific research talent knowledge graph, and there is research-related knowledge feature data between two scientific research talent nodes associated with an entity association link in the basic scientific research talent knowledge graph; The basic scientific research talent knowledge graph is divided according to the previously defined knowledge graph division strategy to generate multiple scientific research talent knowledge sub-graphs.

3. The method for intelligently matching scientific research talents and R&D tasks based on knowledge graph according to claim 1, characterized in that: The determining the matching priority information of each scientific research talent node in the plurality of scientific research talent knowledge sub-graphs according to the plurality of task matching priorities includes: The knowledge entities in the a-th scientific research talent knowledge sub-graph are divided into a first knowledge entity sequence and a second knowledge entity sequence, wherein the knowledge entities in the first knowledge entity sequence and the second knowledge entity sequence do not overlap, and the number of knowledge entities in the first knowledge entity sequence and the second knowledge entity sequence is the same; Constructing a unique mapping relationship between the first knowledge entity sequence and the second knowledge entity sequence, wherein a first knowledge entity in the first knowledge entity sequence and a second knowledge entity in the second knowledge entity sequence have a unique mapping relationship; Sorting and integrating the plurality of task matching priorities to generate a plurality of matching priority integration sequences; Determine a matching priority integration sequence corresponding to each first knowledge entity from the multiple matching priority integration sequences; Determine the matching priority information corresponding to each of the first knowledge entities according to the matching priority integration sequence corresponding to each of the first knowledge entities; The matching priority information corresponding to each of the first knowledge entities is reversely converted to generate matching priority information of the second knowledge entities corresponding to each of the first knowledge entities.

4. The method for intelligently matching scientific research talents and R&D tasks based on knowledge graph according to claim 3 is characterized in that: The method of dividing the knowledge entities in the a-th scientific research talent knowledge sub-graph into a first knowledge entity sequence and a second knowledge entity sequence includes: Obtain the number of knowledge entities in the a-th scientific research talent knowledge sub-graph, where a=1, 2, .., X, where X is the number of scientific research talent knowledge sub-graphs, and X is a positive integer not less than 2; When the number of the knowledge entities is divisible by 2, the knowledge entities in the a-th scientific research talent knowledge sub-graph are all the first knowledge entity sequence and the second knowledge entity sequence; When the number of the knowledge entities is not divisible by 2, a target knowledge entity is randomly extracted from the a-th scientific research talent knowledge sub-graph; The other knowledge entities except the target knowledge entity in the a-th scientific research talent knowledge sub-graph are evenly divided into a first knowledge entity sequence and a second knowledge entity sequence.

5. The method for intelligently matching scientific research talents and R&D tasks based on knowledge graph according to claim 3 is characterized in that: The step of determining the matching priority information corresponding to each of the first knowledge entities according to the matching priority integration sequence corresponding to each of the first knowledge entities includes: Obtaining Y task matching priorities included in the matching priority integration sequence corresponding to the b-th first knowledge entity, where Y is a positive integer not less than 2, b=1, 2, .., Z, where Z is the number of first knowledge entities, and Z is a positive integer not less than 2; The first Y task matching priorities in the matching priority integration sequence corresponding to the bth first knowledge entity are output as matching priorities for associating the first Y candidate R&D tasks to the bth first knowledge entity.

6. The method for intelligently matching scientific research talents and R&D tasks based on knowledge graph according to claim 5 is characterized in that: The reverse conversion of the matching priority information corresponding to each of the first knowledge entities to generate the matching priority information of the second knowledge entities corresponding to each of the first knowledge entities includes: According to the matching priority information corresponding to the b-th first knowledge entity, the first Y candidate R&D tasks are arranged in ascending order of matching priority to generate sorted Y candidate R&D tasks; Arrange the Y task matching priorities in descending order to generate sorted Y task matching priorities; The sorted first Y task matching priorities are output as matching priorities for associating the sorted first Y candidate R&D tasks to the bth second knowledge entity corresponding to the bth first knowledge entity.

7. The method for intelligently matching scientific research talents and R&D tasks based on knowledge graph according to claim 1, characterized in that: The determining the matching priority information of each scientific research talent node in the plurality of scientific research talent knowledge sub-graphs according to the plurality of task matching priorities includes: Annotate the multiple scientific research talent knowledge sub-graphs to generate multiple annotated scientific research talent knowledge sub-graphs, wherein the annotated knowledge points of the two knowledge entities of the first entity association link Guanlan in the annotated scientific research talent knowledge sub-graphs are different; Obtain the number of knowledge points M of knowledge entity annotations in the a-th annotated scientific research talent knowledge sub-graph, where a=1, 2, .., X, where X is the number of scientific research talent knowledge sub-graphs, and X is a positive integer not less than 2; Sorting and integrating the plurality of task matching priorities to generate a plurality of matching priority integration sequences; When M is divisible by 2, obtaining M / 2 matching priority integration sequences from the multiple matching priority integration sequences; Reversely transform the M / 2 matching priority integration sequences to generate M / 2 reversely transformed matching priority integration sequences; Associating the M / 2 matching priority integration sequences and the M / 2 matching priority integration sequences after reverse conversion to the knowledge entities annotated with the M knowledge points, where one knowledge entity annotated with the knowledge point corresponds to one matching priority integration sequence; Determine the matching priority information of each scientific research talent node based on the matching priority integration sequence corresponding to each knowledge entity; The step of labeling the multiple scientific research talent knowledge sub-graphs to generate multiple labeled scientific research talent knowledge sub-graphs includes: Obtaining a set of annotated basic information for annotating a knowledge sub-graph of scientific research talents from a plurality of predetermined data sources, wherein the data sources include a subject classification system library, a terminology standard library in the field of scientific research, and a general research direction guide library in the industry; For each R&D data of the knowledge entity in the scientific research talent knowledge sub-graph, preliminary annotation is performed based on the annotation basic information in the annotation basic information set to generate a scientific research talent knowledge sub-graph set after preliminary annotation; According to the preliminary annotated scientific research talent knowledge sub-graph set and the annotation basic information set, determine the annotation basis of the entity association relationship, generate a relationship annotation basis set, the relationship annotation basis set includes all relationship annotation rules constructed according to the knowledge entity annotation and the annotation basic information, specifically analyze the logical relationship between different knowledge entity annotations, combine the research direction information and subject association rules in the annotation basic information, and construct a relationship annotation rule system; According to the rules in the relationship annotation basis set, the entity association relationships in each of the scientific research talent knowledge sub-graphs are annotated. Specifically, for each pair of knowledge entities with an associated relationship, the annotation content of the associated relationship is determined according to the corresponding entity annotation information and the pre-constructed relationship annotation rules, thereby generating a set of scientific research talent knowledge sub-graphs after relationship annotation as multiple annotated scientific research talent knowledge sub-graphs.

8. The method for intelligently matching scientific research talents and R&D tasks based on knowledge graph according to claim 7, characterized in that: The determining of the matching priority information of each scientific research talent node in the plurality of scientific research talent knowledge sub-graphs includes: When M is not divisible by 2, obtaining (M+1) / 2 matching priority integration sequences from the multiple matching priority integration sequences; Reversely transform (M-1) / 2 matching priority integration sequences in the (M+1) / 2 matching priority integration sequences to generate (M-1) / 2 matching priority integration sequences after reverse transformation; Associating the (M+1) / 2 matching priority integration sequences and the (M-1) / 2 matching priority integration sequences after reverse conversion to the knowledge entities of the M labeled knowledge points, where one knowledge entity labeled with a knowledge point corresponds to one matching priority integration sequence; The matching priority information of each scientific research talent node is determined based on the matching priority integration sequence corresponding to each knowledge entity.

9. The method for intelligently matching scientific research talents and R&D tasks based on knowledge graph according to any one of claims 1 to 8, characterized in that: The method further comprises: Obtaining a research and development task transfer instruction generated by a first interactive port corresponding to the first scientific research talent node, wherein the research and development task transfer instruction includes a characteristic tag of the first research and development task to be transferred and a characteristic tag of the second scientific research talent node; Generate R&D task transfer push information to the second interactive port corresponding to the second scientific research talent node based on the feature tag of the first R&D task and the feature tag of the second scientific research talent node, wherein the R&D task transfer push information includes the feature tag of the first scientific research talent node and the feature tag of the first R&D task; Upon detecting that the second scientific research talent node obtains the first push information of the first research and development task, the research and development task holding information of the first scientific research talent node and the second scientific research talent node is adjusted.

10. A knowledge graph-based intelligent matching system for scientific research talents and R&D tasks, characterized in that: The system for intelligently matching scientific research talents and R&D tasks based on knowledge graphs includes a processor and a memory, wherein the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the method for intelligently matching scientific research talents and R&D tasks based on knowledge graphs as described in any one of claims 1 to 9 above.

Images