A scientific research cooperation recommendation method and system based on a super network

Abstract

The application provides a scientific research cooperation recommendation method and system based on a super network, obtains a multivariate relationship between entities from field-related scientific literature, takes the entities as nodes, takes the multivariate relationship as a super edge, establishes a time label on the edge, forms a multi-dimensional time sequence super network, measures structural similarity between nodes in the multi-dimensional time sequence super network at different times, measures attribute similarity between nodes in the multi-dimensional time sequence super network at different times, fuses the structural similarity and the attribute similarity to calculate a cooperation prediction score between authors, and determines whether to recommend the authors to carry out scientific research cooperation according to the cooperation prediction score. The application can recommend objects with cooperation potential for scientific researchers, can assist selection and configuration of group members, construct an academic cooperation group, and plays an important role in potential cooperation relationship mining and hot research topic discovery.

Description

Technical Field

[0001] This invention relates to the field of intelligent recommendation, and in particular to a research collaboration recommendation method and system based on hypernetworks. Background Technology

[0002] Research collaboration recommendations can be categorized into three types: content-based recommendations, relationship network-based recommendations, and hybrid recommendations. Content-based recommendations typically utilize topic models and vector models to mine academic texts produced by authors, matching and recommending authors with similar content. Rosen et al. introduced author-document relationship information to construct author topic models for author clustering and similar author recommendations. Liu Ping et al. constructed an author-keyword network and divided it into communities, then used the LDA model to calculate the topic similarity of authors within the same community to achieve research collaboration recommendations. Pu Shanshan et al. used author topic models to extract authors' knowledge structure features, then clustered authors based on these features, recommending authors with the highest academic influence in each cluster to form complementary academic teams. Li Zhong et al. used sparse vector models to represent the textual content of authors' papers, achieving better recommendation results compared to the LDA model. Kong et al. constructed a topic-aware vector model, aiming to utilize both author topic information and collaboration information in research collaboration recommendations.

[0003] Recommendation based on relationship networks typically constructs these networks using collaboration, citation, and co-occurrence relationships. Link prediction techniques are then used to assess the likelihood of future collaborations between two non-adjacent nodes (authors). Yu et al. combined Node2vec and LINE models to simultaneously learn the node location and topological features of research collaboration networks, achieving collaboration recommendations through a logistic regression model. Zhou et al. used authors, papers, and keywords as nodes, and collaboration, citation, and inclusion relationships as edges to construct a semantic link network, utilizing semantic paths for collaboration recommendations. Lin et al. extracted information such as authors, institutions, and keywords from scientific literature to construct a co-occurrence network, using network representation learning techniques to identify potential collaborators and institutions for authors.

[0004] Content-based recommendations ignore the relationships between authors, while relationship network-based recommendations do not consider the attribute features of the nodes themselves. Hybrid recommendations combine content-based and relationship network-based recommendations, resulting in better recommendations than either approach. Wang Zhibing et al. defined node institutional attribute similarity and network structure similarity indices in research collaboration networks and weightedly fused these two indices to predict the probability of collaboration between scholars. Zhang Xin et al. used the Node2vec model and the author-topic model to obtain the structural features of the research collaboration network and the topic features of the authors, and then linearly fused these two features to calculate author similarity. Xiong Huixiang et al. used scholar feature words and scholar co-citation relationships to calculate combined similarity to achieve personalized scholar recommendations. Du et al. used attribute network embedding technology to construct a collaboration recommendation model that can simultaneously utilize the structural features of the network and the multidimensional attribute features of scholars. Pradhan et al. used topic models and Doc2vec to extract the academic features of authors, and then used a feature-guided random walk model on the collaboration network to achieve research collaboration recommendations. All of the above methods use ordinary graph structures when constructing relationship networks, which cannot represent the complex and multifaceted relationships between authors, between documents, and between authors and documents. Summary of the Invention

[0005] To address the problems existing in the prior art, a research collaboration recommendation method and system based on hypernetworks is provided, which recommends potential collaborators to researchers by constructing a hypernetwork.

[0006] The first aspect of this invention proposes a research collaboration recommendation method based on hypernetworks, comprising:

[0007] The system extracts multiple relationships between entities from relevant scientific and technological literature, uses entities as nodes and multiple relationships as hyperedges, and establishes time labels on the edges to form a multidimensional temporal hypernetwork.

[0008] A measure of structural similarity between nodes in a multidimensional temporal hypernetwork at different time points;

[0009] Attribute similarity measurement between nodes in a multidimensional temporal hypernetwork at different times;

[0010] For two authors who do not have a collaborative relationship, a collaboration prediction score is calculated by combining structural similarity and attribute similarity. Based on the collaboration prediction score, it is determined whether to recommend the authors to engage in research collaboration.

[0011] As a preferred embodiment, the specific method for forming the multidimensional temporal hypernetwork includes:

[0012] Sentence sets are obtained by text segmentation of massive amounts of scientific and technological literature;

[0013] Given a list of entity types, a large language model is used to determine the matching probability between sentence text features and different entity types. Entity types with a probability less than a preset threshold are removed to obtain the actual list of entities contained in the scientific literature.

[0014] Given a list of multiple relation types, use a large language model to determine and obtain the corresponding relation types of entities in the entity list;

[0015] Different entities are represented as nodes, multi-relationships are represented as edges, and time is used as the feature of the edges to establish author subnets, topic subnets, and institution subnets respectively; the subnets are sliced ​​according to the time window, and hyperedges are established between nodes of different subnets within the same slice to construct a multi-dimensional temporal hypernetwork.

[0016] As a preferred embodiment, in the multidimensional temporal hypernetwork, the correlation matrix H Defined as The matrix is ​​defined with the following elements:

[0017]

[0018] node v With super-edge e If associated, then the matrix H The corresponding element in the code is 1 if it is true, otherwise the corresponding element is 0.

[0019] At this point, the extent of node v at time x is... Defined as:

[0020]

[0021]

[0022] in, To represent a hyperedge e Is it in time? x Appear, For super-edge e Time tags.

[0023] As a preferred embodiment, in the multidimensional temporal hypernetwork, nodes v In the super-edge e The weights on are defined as follows:

[0024]

[0025]

[0026] in, It is a node In scientific and technological literature The weight of a node is assigned based on its frequency of appearance in the literature. It is a node In scientific and technological literature d The order of frequency of occurrence in It is a document d The number of nodes in the middle, Let e ​​be the set of documents corresponding to the hyperedge e.

[0027] As a preferred embodiment, the method for measuring the structural similarity between nodes in the multidimensional temporal hypernetwork at different times includes:

[0028] For any given time point, the weight of the hyperedge corresponding to the node is added to the process of computing the common neighbor index and resource allocation index between nodes;

[0029] The entropy weight method is used to determine the weights between the public neighbor index and the resource allocation index to obtain the structural similarity between nodes at that time point.

[0030] As a preferred embodiment, the method for calculating the public neighbor index is as follows:

[0031]

[0032] The calculation method for resource allocation indicators is as follows:

[0033]

[0034] in, For nodes x and nodes y The public neighbor index at time t, For nodes x and nodes y Resource allocation metrics and parameters at time t a Used to adjust the influence of the node's weight in the hyperedge on the similarity value, if HCN and HRA degenerate into unweighted forms if The metric is equivalent to a regular weighting metric; It is a node x and nodes z The weight at time t is defined as follows:

[0035]

[0036] in, Let be a set of superedges, where each superedge in the set... e All include x and z Two nodes, Indicates the superedge e Is it in time? t Appear, For nodes xIn the super-edge e Weight on, For nodes z In the super-edge e Weights on; These are the weights of nodes z and y at time t, defined similarly.

[0037] As a preferred embodiment, the specific method for calculating the structural similarity is as follows:

[0038]

[0039] in, For nodes x and nodes y Structural similarity at time t; EWM stands for Entropy Weight Method, which uses information entropy to measure the differences within the data of each indicator and determine the weights.

[0040] As a preferred embodiment, the method for measuring the attribute similarity between nodes in the multidimensional temporal hypernetwork at different times includes:

[0041] For any given time point, the nodes are mapped into vector form using a dynamic agent model, and then input into a variational autoencoder to generate the node's attribute feature vectors in an unsupervised manner.

[0042] Calculate the Euclidean distance between the attribute feature vectors of the nodes to obtain the attribute similarity between the nodes at that time point.

[0043] As a preferred embodiment, the method for calculating the cooperative prediction score is as follows:

[0044]

[0045]

[0046]

[0047] in, For the author x, y Collaborative prediction scores and Each represents the author x, y Overall structural similarity and overall attribute similarity, , Each represents the author x, y In a multidimensional temporal hypernetwork, the corresponding author nodes, topic nodes, and institution nodes, the balancing parameters are... Used to adjust the structural similarity value and the weight of the similarity values. Each represents the author x With the author y Structural similarity among author nodes, topic nodes, and structure nodes. Each represents the author x With the author y The similarity of attributes between author nodes, topic nodes, and structure nodes.

[0048] The second aspect of this invention proposes a research collaboration recommendation system based on a hypernetwork, comprising a computer system and a computer program, wherein the computer program executes the method described in the first aspect when run by the computer system. Compared with the prior art, the beneficial effects of adopting the above technical solution are as follows:

[0049] (1) This invention can recommend potential partners for researchers, assist in the selection and allocation of group members, build academic cooperation groups, and play an important role in the discovery of potential cooperative relationships and hot research topics.

[0050] (2) This invention proposes to model the multi-relationship between objects based on scientific research cooperation hypernetwork, so as to avoid information loss caused by the conversion of multi-relationship to binary relationship.

[0051] (3) This invention utilizes both structural similarity features and attribute similarity features on the research collaboration hypernetwork to achieve hybrid recommendation of research collaboration, which avoids the problem of sparse explicit collaboration relationships and the problem of overly concentrated and undiversified implicit collaboration recommendation results. Attached Figure Description

[0052] Figure 1 This is a flowchart of the research collaboration recommendation method based on hypernetworks proposed in this invention.

[0053] Figure 2 This is a schematic diagram of a multidimensional temporal hypernetwork constructed in one embodiment of the present invention.

[0054] Figure 3 This is a schematic diagram illustrating the generation of attribute feature vectors in one embodiment of the present invention. Detailed Implementation

[0055] The embodiments of this application are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar modules or modules having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application. Rather, the embodiments of this application include all variations, modifications, and equivalents falling within the spirit and scope of the appended claims.

[0056] In the era of big data, where information overload is increasingly severe, how to help researchers find the information they need from massive amounts of academic data has become an urgent problem to be solved. Search engines can assist researchers in filtering and selecting information from the sea of ​​data, but they can only passively obey input commands and cannot proactively recommend content of interest to users. The research collaboration recommendation method proposed in this invention can solve the problem of information overload while providing personalized recommendation services to users, offering reference and suggestions for the selection and allocation of team members. Please refer to... Figure 1 The specific plan is as follows:

[0057] Step 1: Construct a multidimensional temporal hypernetwork.

[0058] In this embodiment, a large oracle model is used to extract multivariate relationships between entities from domain-related scientific and technological literature. Entities are treated as nodes, and multivariate relationships as hyperedges, with time labels added to the edges to form a multidimensional temporal hypernetwork. In one embodiment, entities such as authors, topics, and institutions can be extracted from scientific and technological literature. The construction of the multidimensional temporal hypernetwork mainly includes two parts: entity recognition and multivariate relationship extraction, and hypernetwork construction. These two parts are described in detail below.

[0059] (1) Entity recognition and multivariate relation extraction

[0060] Scientific and technological literature d Text segmentation is performed to obtain a sentence set. S Then, for each sentence in turn. Construct prompts to obtain entities and multiple relationships through two rounds of question-and-answer interaction with a large language model.

[0061] First round of Q&A:

[0062] Given a list of entity types, use a large language model to determine the matching probability between sentence text features and different entity types. A suggested approach is as follows:

[0063] Given the list of entity types: [Entity type 1, Entity type 2, ...]

[0064] Question: What entity types might be included in the sentence "text...", and give the probability.

[0065] Entity types with a probability less than a preset threshold are removed to obtain the actual list of entities contained in the scientific and technological literature (entity types). s1 Entity type s2 , ..., entity type sn ).

[0066] Second round of Q&A:

[0067] Given a list of multiple relation types, use a large language model to determine and obtain the corresponding relation types of entities in the entity list. The suggested construction is as follows:

[0068] The given list of multi-relation types is: [relation type 1, relation type 2, ...]

[0069] Question: Given the sentence "text...", determine which of the following entities are in the given sentence: (entity type s1, entity type s2, ..., entity type...). The relation type corresponding to sn).

[0070] The entity and the multiple relationships between entities can be obtained through two rounds of question and answer.

[0071] (2) Construction of multidimensional temporal hypernetwork

[0072] Please refer to Figure 2 This study performs entity recognition and multivariate relation extraction on massive amounts of scientific and technological literature, representing different entities as nodes and multivariate relations as edges, with time as a feature of the edges, and establishes author subnetworks, topic subnetworks, and institution subnetworks respectively. Then, the subnetworks are sliced ​​according to time windows, and hyperedges are established between nodes of different subnetworks within the same slice to construct a multidimensional temporal hypernetwork.

[0073] In this embodiment, a multidimensional temporal hypernetwork is defined as... ,in, It is a finite set of nodes. n It is the number of nodes; It is a set of node attributes. Represents a node Corresponding text attributes; It is a super-edge set. It is the number of superedges; Indicates the super edge. m It is the number of nodes in the hyperedge. ; It is a super-edge e Time tags, It is a node v In the super-edge e Weights on the network. Correspondence matrix of the hypernetwork. H It is The matrix is ​​defined with the following elements:

[0074]

[0075] node v With super-edge e If associated, then the matrix H The corresponding element in the hypernetic network has a value of 1 if it is true and 0 otherwise. v In time x The salvation Defined as:

[0076]

[0077]

[0078] in, To represent a hyperedge e Is it in time? x Appear, For super-edge e Time tags.

[0079] Each hyperedge contains multiple nodes, and the contributions of different nodes to the hyperedge are often different, requiring the appropriate allocation of weights to the nodes. This embodiment also proposes a node... v In the super-edge e The formula for calculating the weights on the surface is as follows:

[0080] (1)

[0081] (2)

[0082] in, It is a node In the literature The weight of a node is assigned based on its frequency of appearance in the literature. It is a node In the paper d The order of frequency of occurrence in It is a document d The number of nodes in the middle, Let e ​​be the set of documents corresponding to the hyperedge e. It can be seen that formula (2) can effectively assign weights to nodes, satisfying the following three principles: 1) In the papers... d 1) The more frequently a node appears in a node, the greater its weight; 2) Each node receives a reasonable weight; 3) In the nodes... v With the ranking unchanged, the larger the number of nodes, the more nodes... v The smaller the weight.

[0083] Step 2: Measuring the structural similarity between nodes in a multidimensional temporal hypernetwork at different times.

[0084] Traditional structural similarity metrics are implemented by studying the topology of ordinary networks and have achieved high accuracy on publicly available datasets, such as Common Neighbor (CN) and Resource Allocation (RA). Among these, the CN metric considers the similarity between two nodes... x and y The more common neighbors two nodes have, the more similar the two nodes are, as defined in formula (3):

[0085] (3)

[0086] The RA metric considers the similarity between two nodes to be determined by a single node.x Pass to another node y The amount of resources is determined by the quantity of resources, as defined in formula (4):

[0087] (4)

[0088] It is a node x The set of neighboring nodes, Neighboring nodes z Degree. Node x Neighbors act as intermediaries for resource transfer. Each intermediary also possesses one unit of resource, which it distributes equally among its neighbors. The final node... y The amount of resources received is the node x and nodes y The similarity value.

[0089] The quality of structural similarity metrics depends on their ability to accurately capture the structural features of a network. CN and RA metrics only consider binary relationships between nodes in a typical graph and ignore the influence of node and edge weights on similarity. In fact, the proper use of weight information in hypernetworks can effectively improve the accuracy of link prediction. Therefore, this embodiment adds the weights of the hyperedges corresponding to nodes to the calculation of the common neighbor metric and resource allocation metric between nodes, making them applicable to multidimensional temporal hypernetworks to achieve structural similarity measurement. Specifically, this embodiment proposes a specific definition method for the common neighbor metric HCN and resource allocation metric HRA of multidimensional temporal hypernetworks, as shown in formulas (5) and (6):

[0090] (5)

[0091] (6)

[0092] Among them, parameters a This is used to adjust the impact of the node's weight in the hyperedge on the similarity value. If HCN and HRA degenerate into unweighted forms; if The metric is equivalent to a regular weighting metric. is the weight of the corresponding hyperedge of node x and node z, defined as shown in formula (7):

[0093] (7)

[0094] Among them, hyper-edge set Each hyperedge e in the set contains two nodes, x and z. Indicates the superedge e Is it in time? t Appear, For nodesx In the super-edge e Weight on, For nodes z In the super-edge e Weights on; These are the weights of nodes z and y at time t, defined similarly.

[0095] HCN measures the similarity of the neighbor structure of hypernetwork nodes, while HRA calculates the amount of resources transferred between two hypernetwork nodes. The higher the values ​​of these two metrics, the more similar the local structures of the nodes are. According to the principle of clustering, two nodes with similar local structures are more likely to form an edge.

[0096] In this embodiment, the HCN and HRA indices obtained jointly by the entropy weight method are used to measure the structural similarity between nodes. Specifically, the structural similarity between node x and node y at time t is calculated as follows:

[0097]

[0098] Here, EWM (the entropy weight method) refers to the entropy weight method, which uses information entropy to measure the differences within the data of each indicator and determine the weights. It should be noted that structural similarity measurement is calculated within the subnet, that is, the measurement is performed between nodes within the author subnet, between nodes within the topic subnet, and between nodes within the institution subnet.

[0099] Step 3: Measure the attribute similarity between nodes in a multidimensional temporal hypernetwork at different times.

[0100] In constructing relational networks, the problem of network sparsity caused by a lack of literature is often encountered, making it difficult to obtain accurate collaborative recommendation results by relying solely on topology. Existing recommendation research shows that node attributes can improve the accuracy of recommendation models and alleviate the cold start and data sparsity problems. Nodes in multidimensional temporal hypernetworks carry a large number of textual attributes, and semantic analysis of these attributes can effectively uncover implicit relationships between nodes, compensating for the inadequacy of topological similarity measurements. Topic models are a key technology in text semantic analysis and are widely used in predicting scientific research collaborations, detecting domain hotspots, and discovering communities. Dynamic Topic Models (DTMs) introduce information about the evolution of topics over time based on Latent Dirichlet Allocation (LDA).

[0101] In this embodiment, a dynamic topic model is used to construct the attribute features of nodes, transforming attribute similarity calculation into distance calculation between vectors. The output of the DTM is a text-topic matrix. Topic-Word Matrix Time-Topic Matrix Using matrices and Time can be set t The node of time v Mapped to vector form:

[0102]

[0103] The dimension of the vector is the same as the number of topics and the value of each dimension is non-negative. Directly using it as the attribute feature vector of a node will affect the flexibility and computability of the vector representation.

[0104] Furthermore, this embodiment employs a variational autoencoder (SAE) to model node features, generating attribute feature vectors in an unsupervised manner. The variational autoencoder structure is as follows: Figure 3 As shown, where, It is the input of the encoder, and the output is a variational distribution. , It is a latent vector. These are the parameters of the inference network. The input to the decoder is... The output is a probability distribution p , These are the parameters for generating the network. Let... When input into the encoder, the latent vector z becomes the constructed attribute feature vector.

[0105] After obtaining the attribute feature vectors of the nodes, this embodiment uses Euclidean distance to evaluate the nodes. x and y The similarity of attributes between them is defined as shown in formula (10):

[0106]

[0107] in, For time node t x, y Similarity of attributes between them.

[0108] It should be noted that attribute similarity measurement is calculated within subnets, that is, between nodes within the author subnet, between nodes within the topic subnet, and between nodes within the institution subnet.

[0109] Step 4: Recommendation of scientific research collaborations.

[0110] In scientific research, collaborative relationships among authors are categorized into two types: explicit and implicit. Explicit relationships are established through co-authored publications, while implicit relationships refer to potential collaborations formed by authors who, despite not directly collaborating due to factors such as location and time, share similar research interests and themes. Traditional collaborative recommendation algorithms typically use either explicit or implicit relationships to construct recommendation models. The former faces the problem of sparse explicit collaborations, while the latter can lead to overly concentrated and undiversified recommendation results, resulting in poor recommendation performance.

[0111] In this embodiment, the structural similarity index and attribute similarity index represent the explicit and implicit relationships in the multidimensional temporal hypernetwork, respectively. To capture the features of both types of cooperative relationships simultaneously in cooperative recommendation, the two metrics are fused to predict the author. x With the author y In time t The probability of cooperation is predicted, and the score is calculated using the following formula:

[0112]

[0113]

[0114]

[0115] in, For the author x, y Collaborative prediction scores and Each represents the author x, y Overall structural similarity and overall attribute similarity, , Each represents the author x, y In a multidimensional temporal hypernetwork, the corresponding author nodes, topic nodes, and institution nodes, the balancing parameters are... It is used to adjust the structural similarity value and the weight of the similarity value. The smaller the absolute value, the greater the impact of the structural similarity of the hypernetwork on the recommendation results; The larger the absolute value, the stronger the dominant role of attribute similarity between nodes in collaborative recommendation. These represent the authors at time node t. x With the author y Structural similarity among author nodes, topic nodes, and structure nodes. These represent the authors at time node t. x With the author y The similarity of attributes between author nodes, topic nodes, and structure nodes.

[0116] For two authors who have no collaborative relationship x, y The calculated predicted scores The larger the value, the stronger the author's... x and the author y The higher the probability of establishing a partnership, the greater the likelihood of success; when the predicted score exceeds the recommendation threshold, the author is recommended. x and the author y To conduct scientific research collaborations.

[0117] This embodiment also proposes a research collaboration recommendation system based on hypernetworks, including a computer system and a computer program. When the computer program is run by the computer system, it executes the aforementioned research collaboration recommendation method based on hypernetworks.

[0118] In particular, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts.

[0119] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such transmitted data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0120] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0121] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0122] In another aspect, this application also provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the hypernetwork-based research collaboration recommendation method described in the above embodiments.

[0123] In another aspect, this application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to implement the hypernetwork-based research collaboration recommendation method described in the above embodiments.

[0124] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0125] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the methods according to the embodiments of this application.

[0126] For those skilled in the art, the specific meanings of the above terms in this invention can be understood according to the specific circumstances; the accompanying drawings in the embodiments are used to clearly and completely describe the technical solutions in the embodiments of this invention. Obviously, the described embodiments are some embodiments of this invention, but not all embodiments. Generally, the components of the embodiments of this invention described and shown in the accompanying drawings can be arranged and designed in various different configurations.

[0127] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1.A method for recommending scientific research cooperation based on a super network, characterized in that, include: The system extracts multiple relationships between entities from relevant scientific and technological literature, uses entities as nodes and multiple relationships as hyperedges, and establishes time labels on the edges to form a multidimensional temporal hypernetwork. A measure of structural similarity between nodes in a multidimensional temporal hypernetwork at different time points; Attribute similarity measurement between nodes in a multidimensional temporal hypernetwork at different times; For two authors who do not have a collaborative relationship, a collaboration prediction score is calculated by combining structural similarity and attribute similarity. Based on the collaboration prediction score, it is determined whether to recommend the authors to conduct research collaboration. The methods for measuring the structural similarity between nodes in a multidimensional temporal hypernetwork at different times include: For any given time point, the weight of the hyperedge corresponding to the node is added to the process of computing the common neighbor index and resource allocation index between nodes; By using the entropy weight method to determine the weights between the common neighbor index and the resource allocation index, the structural similarity between nodes at this point in time can be obtained. The calculation method for the public neighbor index is as follows: The calculation method for resource allocation indicators is as follows: where is the node x and the node y is the resource allocation indicator at time t t for the node and the node x is the similarity indicator at time t for the node y and the node a is a parameter used to adjust the influence of the weight of a node in a hyperedge on the similarity value, if , the HCN and HRA degenerate to the unweighted form, if , the metric indicator is equivalent to the ordinary weighted indicator; is the weight of the node x and the node z at time t, defined as follows: wherein, is a hyperedge, each hyperedge in the set e contains x and z two nodes, denotes whether the hyperedge e occurs at time t , is the weight of node x on hyperedge e , is the weight of node z on hyperedge e ; is the weight of node z and node y at time t, the definition is the same. 2.The method of claim 1, wherein, The specific methods for forming the multidimensional temporal hypernetwork include: Sentence sets are obtained by text segmentation of massive amounts of scientific and technological literature; Given a list of entity types, a large language model is used to determine the matching probability between sentence text features and different entity types. Entity types with a probability less than a preset threshold are removed to obtain the actual list of entities contained in the scientific literature. Given a list of multiple relation types, use a large language model to determine and obtain the corresponding relation types of entities in the entity list; Different entities are represented as nodes, multi-dimensional relationships are represented as edges, and time is used as the feature of the edges to establish author subnets, topic subnets, and institution subnets respectively; the subnets are sliced ​​according to the time window, and hyperedges are established between nodes of different subnets within the same slice to construct a multi-dimensional temporal hypernetwork. 3.The method of claim 2, wherein, In the multi-dimensional time-sequential hypernetwork, the incidence matrix H is defined as a matrix whose elements are defined as follows: node v With super-edge e If associated, then the matrix H The corresponding element in the code is 1 if it is true, otherwise the corresponding element is 0. At this time, the node v At time x of the super degree is defined as: wherein, is a hyperedge e whether a hyperedge x occurs at a time is a time label of a hyperedge e . 4.The method of claim 3, wherein, In the multidimensional temporal hypernetwork, the weight of node v on hyperedge e is defined as follows: wherein, is the node in the scientific literature occupies the weight, the weight is assigned by the frequency of occurrence of the node in the literature, is the node in the scientific literature d in the order of the frequency of occurrence, is the number of nodes d in the literature is the set of documents corresponding to the hyperedge e. 5.The method of claim 1, wherein, The specific method for calculating the structural similarity is as follows: wherein is a node x and a node y the structural similarity at time t; EWM represents the entropy weight method, which measures the difference in each index data by information entropy and determines the weight. 6.The method of claim 1 or 2, wherein, The methods for measuring the attribute similarity between nodes in a multidimensional temporal hypernetwork at different times include: For any given time point, the nodes are mapped into vector form using a dynamic agent model, and then input into a variational autoencoder to generate the node's attribute feature vectors in an unsupervised manner. Calculate the Euclidean distance between the attribute feature vectors of the nodes to obtain the attribute similarity between the nodes at that time point. 7.The method of claim 6, wherein, The method for calculating the cooperative prediction score is as follows: wherein, is the author x , y , , is the author x , y , , is the author x , y , is a balance parameter used to adjust the proportion of the structure similarity value and the similarity value, is the structure similarity between the author node, the topic node and the structure node of the author x and the author y at time node t, is the attribute similarity between the author node, the topic node and the structure node of the author x and the author y at time node t. 8.A scientific research cooperation recommendation system based on a super network, characterized in that, It includes a computer system and a computer program, wherein the computer program executes the method as described in any one of claims 1 to 7 when the computer system is running.

Images